oEPA
United States
Environmental Protection
Agency
Technical Guidance for Constructing
a Human Well-being Index (HWBI):
A Puerto Rico Example
National Health arid Environmental Effects Research Laboratory
Office of Research and Development
EPA/600/R-16/363
April 2017
http://www.epa.gov/si
-------
EPA/600/R-16/363
April 2017
Technical Guidance for
Constructing a
Human Well-being Index (HWBI):
A Puerto Rico Example
By
Jessica L. Orlando, Susan H. Yee, Linda C. Harwell, and Lisa M. Smith
Gulf Ecology Division
National Health and Environmental Effects Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Gulf Breeze, FL 32561
i
-------
Notice and Disclaimer
The U.S. Environmental Protection Agency through its Office of Research and Development
(ORD) funded and collaborated in the research described herein. This document has been
subjected to the Agency's peer and administrative review and has been approved for
publication as an EPA document. Any mention of trade names, products, or services does not
imply an endorsement or recommendation for use.
This is a contribution to the EPA ORD Sustainable and Healthy Communities Research Program.
The appropriate citation for this report is:
Orlando, J.L., S.H. Yee, L.C. Harwell, and L.M. Smith. 2017. Technical Guidance for Constructing
a Human Well-being Index (HWBI): A Puerto Rico Example. U.S. Environmental Protection
Agency, Gulf Breeze, FL, EPA/600/R-16/363.
Acknowledgments
We greatly appreciate the efforts of peer reviewers who took the time to read the report and
evaluate the usability of the guidance and code: Kyle Buck (US EPA) and Kasey Jacobs (US Forest
Service, International Institute of Tropical Forestry). Angelica Sullivan provided an editorial
review and assistance with formatting.
Cover photo credits: Stephanie Orlando and Donald Yee
ii
-------
Table of Contents
Notice and Disclaimer ii
Acknowledgments ii
Table of Contents Error! Bookmark not defined.
List of Tables v
List of Figures viii
List of Code Boxes xi
Executive Summary xii
Chapter 1: Introduction 1
1.1. Background 1
1.2. Adapting the HWBI to Puerto Rico 4
Chapter 2: Identifying Metrics and Finding Data 7
2.1. HWBI Structure 7
2.2 Metric Selection for Puerto Rico HWBI 7
2.3. Substitute Metrics 12
2.4. Spatial Availability of Metric Data 14
Chapter 3: Downloading and Organizing Data 16
3.1. Downloading HWBI Data 16
3.2. Spatial and Temporal Tradeoff 17
3.3. Organizing and Formatting Downloaded Metric Data 17
Example 1: Organizing and formatting metric data based on multiple spatial scales 17
Example 2: Organizing and formatting metric data based on multiple variables 21
Example 3: Organizing and formatting metric data based on occurrence data 23
3.4. Merging Metric Data by municipio x year Combination 26
3.5. Selecting the Finest Available Spatial Scale for Each metricxcountyxyear Combination. 27
Example 4: Selecting finest spatial scale for metric data based on multiple spatial scales. 31
Example 5: Selecting finest spatial scale for metric data based on more complex spatial
scales 31
Chapter 4: Processing Data for HWBI Calculation 35
4.1. Summary of Data Processing Prior to HWBI Calculation 35
4.2. Processing Metric Data Prior to HWBI Calculation 36
Processing step 1: Correct data structure and temporal imputation 37
Step la. Construct a county-level metric dataset (metric x county x year) 37
Step lb. Identify original metric data source and populate template 39
Step lc. Temporally impute missing data 42
Processing step 2: Create fill data for spatial imputation 45
iii
-------
Processing step 3: Complete metric data with spatial imputations 47
Processing step 4: Normalize and standardize metric values 52
Step 4a. Normalize metric values 52
Step 4b. Standardize metric values 52
Chapter 5: Calculating HWBI 56
5.1. Summary of HWBI Calculation 56
5.2. Calculating HWBI 57
Processing step 1: Mean decadal (2000-2010) indicator score calculation by municipio... 57
Processing step 2: Mean decadal (2000-2010) domain score calculation by municipio 58
Processing step 3: Mean decadal (2000-2010) HWBI calculation by municipio 59
5.3. Spatial and Temporal Flexibility in Calculation 59
Example 1: Annual HWBI calculation by municipio 60
Example 2: Mean decadal (2000-2010) HWBI calculation by commonwealth (state
equivalent) 61
Chapter 6: Mapping HWBI 62
6.1. Mapping HWBI in R 62
6.2. Maps of Puerto Rico HWBI 65
Chapter 7: Comparing HWBI Across Time 75
7.1. Summary of Temporal Flexibility of HWBI 75
7.2. Evaluation of HWBI Time Series 75
Chapter 8: Comparing HWBI Among Populations 81
8.1. Considerations for HWBI Comparison Among Populations 81
8.2. Evaluation of Alternative Methods for Missing Metrics 81
8.3. Comparison of Puerto Rico and U.S. HWBI 89
8.4. Comparison of HWBI among San Juan Bay Estuary Watershed Municipios 95
Chapter 9: Summary 99
References 100
Appendix A: HWBI Metrics 105
Appendix B: Variables in HWBI Processing 130
iv
-------
List of Tables
Table 1.1. Descriptions of domains in the Human Well-being Index 3
Table 2.1. Summary of 80 metrics composing 25 indicators within 8 domains of well-
being 9
Table 2.2. Summary of metric availability for Puerto Rico compared to U.S. HWBI 11
Table 3.1. Unformatted and unprocessed 2012 median home value data (U.S. dollars;
HD01_VD01) by county for available municipios (GEO.id2) downloaded from U.S.
Census Bureau, American Community Survey using the web-based database
American FactFinder 19
Table 3.2. Sample of formatted county-level metric data for 2012 median home value
(U.S. dollars) by municipio for available municipios (HOMEVAL_m) 19
Table 3.3. Unformatted and unprocessed 2012 median home value data (U.S. dollars;
HD01_VD01) by county for available Public Use Microdata Areas (PUMAs; GEO.id2)
downloaded from U.S. Census Bureau, American Community Survey using the web-
based database American FactFinder 20
Table 3.4. Sample of formatted county-level metric data for 2012 median home value
(U.S. dollars) by municipio for available municipios (HOMEVAL_m) and Public Use
Microdata Areas (PUMAs; HOMEVAL_plO) using 2010 U.S. Census boundaries 20
Table 3.5. Sample of unformatted and unprocessed 2012 voter registration data
(Electores) by Senate District and Precinct (Precinto) downloaded from Comision
Estatal de Elecciones 21
Table 3.6. Sample of formatted county-level metric data for 2012 voter registration by
municipio 23
Table 3.7. Sample of unformatted and unprocessed diabetes mortality occurrence data
downloaded from the Centers for Disease Control and Prevention-National Center
for Health Statistics' National Vital Statistics System Multiple Cause of Death
database 24
Table 3.8. Sample of formatted county-level metric data for 2000 diabetes mortality
(percentage) by municipio for available spatial scales, including municipio
(DIABMORTjm), state without county-level data (DIABMORT_PR.m), MSA
(DIABMORTjmsa), and state without MSA-level data (DIABMORT_PR.msa) 26
Table 3.9. Sample of formatted county-level metric data for San Juan municipio merged
by municipio and year for available spatial scales 27
Table 3.10. Sample of wide-format county-level median home value (HOMEVAL) metric
data for all available spatial scales 29
v
-------
Table 3.11. Sample of wide-format county-level diabetes mortality (DIABMORT) metric
data for all available spatial scales 30
Table 3.12. Sample of long-format county-level metric data including data for all
available spatial scales 31
Table 3.13. Sample of output data with finest available spatial scale selected for each
metric x county x year combination for the densely populated San Juan municipio
and less populated Adjuntas municipio 34
Table 4.1. Summary of HWBI versions for Puerto Rico using alternative processing
methods 35
Table 4.2. Sample of lookup table for Puerto Rico spatial scales within spatial
hierarchy 37
Table 4.3. Sample of lookup table of HWBI metric descriptions 37
Table 4.4. Sample of county-level metric dataset for Puerto Rico where each row is a
unique metric x county x year combination 38
Table 4.5. Sample of output for finest available spatial scale by metric, county, and
year 41
Table 4.6. Example of temporal imputation where missing metric data (measure) were
forward filled with data from the previous year 44
Table 4.7. Example of temporal imputation where missing metric data (measure) were
backward filled with data from the following year 44
Table 4.8. Example of temporal imputation where missing metric data (measure) were
both forward and backward filled with data from previous and following years 45
Table 4.9. Sample step 1 output where original data source is identified (ORIG_FIPS) and
missing data have been temporally imputed 45
Table 4.10. Sample step 2 output (Domain Metrics Fill (RUCC-GINI) used for spatial
imputation) 47
Table 4.11. Sample step 3 output (metrics completed with spatial imputation) 51
Table 4.12. Sample step 4 output (metrics normalized and standardized) 54
Table 5.1. Sample output of Puerto Rico mean decadal (2000-2010) HWBI indicator
scores by municipio (unique Federal Information Processing Standards [FIPS]
code) 58
Table 5.2. Sample output of Puerto Rico mean decadal (2000-2010) HWBI domain
scores by municipio (unique Federal Information Processing Standards [FIPS]
code) 58
Table 5.3. Sample output of Puerto Rico mean decadal (2000-2010) HWBI by municipio
(unique Federal Information Processing Standards [FIPS] code) 59
vi
-------
Table 5.4 Sample of Puerto Rico HWBI by year 60
Table 5.5. Puerto Rico mean decadal (2000-2010) HWBI 61
Table 6.1. Sample of fortified data frame of polygons joined with shapefile attributes 62
Table Al. Summary of metrics used for HWBI calculation for Puerto Rico, organized by
domain (bold) and indicator (italics) 105
Table Bl. Description of HWBI processing output for Processing step 1: Correct data
structure and temporal imputation 130
Table B2. Description of HWBI processing output for Processing step 2: Create metric fill
data for spatial imputation 132
Table B3. Description of HWBI processing output for Processing step 3: Complete metric
data with spatial imputations 133
Table B4. Description of HWBI processing output for Processing step 4: Normalize and
standardize metric values 136
vii
-------
List of Figures
Figure 1.1. Conceptual diagram linking economic, social, and environmental flows of
goods and services to the HWBI 2
Figure 1.2. Map of Puerto Rico in relation to contiguous United States 4
Figure 1.3. Map of Puerto Rico including boundaries of 78 municipios 5
Figure 2.1. Diagrammatic representation of human well-being components 8
Figure 2.2. Decision tree used to select metric data for Puerto Rico adaptation of HWBI.. 13
Figure 2.3. Maps of Puerto Rico with boundaries for HWBI metric data spatial scales
including by a) Commonwealth of Puerto Rico (state equivalent), b) municipio
(county equivalent), c) Public Use Microdata Area (PUMA) based on 2000 U.S.
Census, d) PUMA based on 2010 U.S. Census, e) Metropolitan Statistical Area (MSA)
based on 2000 U.S. Census, f) MSA based on 2010 U.S. Census, and g) World Values
Survey region 14
Figure 3.1. Sample screen-shot of U.S. Census Bureau's web based interface for
downloading U.S. Census Data (www.factfinder.census.gov) 18
Figure 3.2. Summary of spatial hierarchy used to select data for Puerto Rico HWBI 28
Figure 4.1. Summary of processing steps to prepare data for HWBI calculation 36
Figure 4.2. Summary of spatial imputation of metrics for Puerto Rico HWBI 48
Figure 5.1. Summary of HWBI calculation methods for indicator score, domain score,
and composite index 56
Figure 6.1. Map of Puerto Rico plotted from fortified shapefile in R 63
Figure 6.2. Maps of mean decadal (2000 - 2010) HWBI for Puerto Rico by municipio
where HWBI was calculated a) to compare HWBI within Puerto Rico (within Puerto
Rico) and b) to compare HWBI in Puerto Rico to the rest of the U.S. (U.S. imputations
for only missing indicators) 66
Figure 6.3. Maps of mean decadal (2000-2010) HWBI domain scores for Puerto Rico by
municipio where HWBI was calculated a) to compare HWBI within Puerto Rico
(within Puerto Rico) and b) to compare HWBI in Puerto Rico to the rest of the U.S.
(U.S. imputations for only missing indicators) 67
Figure 6.4. Maps of mean decadal (2000 - 2010) HWBI indicator scores for Puerto Rico
by municipio where HWBI was calculated a) to compare HWBI within Puerto Rico
(within Puerto Rico), and b) to compare HWBI in Puerto Rico to the rest of the U.S.
(U.S. imputations for only missing indicators) 69
viii
-------
Figure 7.1. Time series for Puerto Rico HWBI (U.S. imputations for only missing
indicators) 76
Figure 7.2. Net change in Puerto Rico HWBI by municipio for 2000 - 2013 (U.S.
imputations for only missing indicators) 76
Figure 7.3. Time series for Puerto Rico HWBI by municipio (U.S. imputations for only
missing indicators) 77
Figure 7.4. Time series for Puerto Rico HWBI indicator scores (U.S. imputations for only
missing indicators) 79
Figure 8.1. Comparison of mean decadal (2000-2010) metric values for Puerto Rico
calculated using substituted measures and U.S. imputations 82
Figure 8.2. Comparison of substituted and imputed metrics for Puerto Rico 82
Figure 8.3. Comparison of mean decadal (2000-2010) HWBI for Puerto Rico calculated
with substituted (72 metrics) and without substituted measures (60 metrics with
U.S. imputations only for missing indicators) 83
Figure 8.4. Comparison of unstandardized metric values for Puerto Rico, Hawai'i, and
the United States 85
Figure 8.5. Comparison of mean decadal (2000-2010) HWBI for Puerto Rico using
alternative methods based on number of imputations (all missing metrics or only
when a missing metric resulted in a missing indicator) and source of imputations
(Hawai'i by RUCC or U.S. by RUCC-GINI) 87
Figure 8.6. Comparison of mean decadal (2000-2010) domain scores for Puerto Rico
using alternative methods based on number of imputations (all missing metrics or
only when a missing metric resulted in a missing indicator) and source of
imputations (Hawai'i by RUCC or U.S. by RUCC-GINI) 88
Figure 8.7. Comparison of mean decadal (2000 - 2010) HWBI for U.S states and Puerto
Rico 89
Figure 8.8. Mean decadal (2000-2010) HWBI percentile rankings for 78 municipios of
Puerto Rico (U.S. imputations for only missing indicators) compared to U.S. county
equivalents (n=3,221) 90
Figure 8.9. Annual time series (2000-2010) of HWBI values for U.S. states, Washington,
D.C., and Puerto Rico (U.S. imputations for only missing indicators) 91
Figure 8.10. Comparison of mean decadal (2000-2010) domain scores for Puerto Rico
(U.S. imputations for only missing indicators) and the U.S. states with highest (New
Hampshire) and lowest (Louisiana) HWBIs 92
Figure 8.11. Comparison of mean decadal (2000-2010) indicator scores for Puerto Rico
(U.S. imputations for only missing indicators) and the U.S. states with highest (New
Hampshire) and lowest (Louisiana) HWBIs 93
ix
-------
Figure 8.12. Map of a) Puerto Rico's San Juan Bay Estuary Watershed and b) eight
associated municipios 95
Figure 8.13. Mean decadal (2000-2010) HWBI percentile rankings for 78 municipios of
Puerto Rico (U.S. imputations for only missing indicators) compared to U.S. county
equivalents (n=3,221) 96
Figure 8.14. Puerto Rico mean decadal (2000-2010) HWBI indicator scores for eight
municipios of San Juan Bay Estuary Watershed (U.S. imputations for only missing
indicators) 97
x
-------
List of Code Boxes
Box 3.1. R Code to read in municipio-level HOMEVAL data for 2012 19
Box 3.2. R Code to format county-level metric data for 2012 median home value 19
Box 3.3. R Code to read in PUMA-level HOMEVAL data for 2012 and apply data to
PUMA 2010 geography 20
Box 3.4. R Code to read in 2012 data based on senatorial district maps 21
Box 3.5. R Code to calculate percent of population registered to vote 22
Box 3.6. R Code to read in diabetes mortality occurrence data for 2000 23
Box 3.7. R Code to calculate percentage of deaths attributed to diabetes 24
Box 3.8. R Code to reformat data from wide to long 30
Box 3.9. R Code to select data at the finest available spatial scale 32
Box 4.1. R Code to load input data and get unique metric variables with data available
for Puerto Rico for each municipio and year 38
Box 4.2. R Code to subset data by finest available spatial scale and join to the county-
level metric template 39
Box 4.3. R Code to temporally impute missing data with data from previous or
following year 42
Box 4.4. R Code to group data by metric, year, and RUCC-GINI combination in
preparation for spatial imputation 46
Box 4.5. R Code to impute missing data based on RUCC-GINI code 48
Box 4.6. R Code to normalize metric values 52
Box 4.7. R Code to standardize metric values between 0.1 and 0.9 53
Box 5.1. R Code to calculate indicator scores as arithmetic mean of metrics, by
municipio and weighted population across years 57
Box 5.2. R Code to calculated domain scores as arithmetic means of indicator scores
by municipio 58
Box 5.3. R Code to calculated composite HWBI as geometric mean of domain scores for
each municipio 59
Box 5.4. R Code to calculate HWBI annually for each municipio 60
Box 5.5. R Code to calculate mean decadal HWBI by "state" or equivalent (i.e., Puerto
Rico) 61
Box 6.1. R Code to merge attributes with polygons in shapefile 62
Box 6.2. R Code for customized mapping of HWBI by county (Fig. 6.2), by domain (Fig.
6.3), and by indicator (Fig. 6.4) 63
xi
-------
Executive Summary
The U.S. Environmental Protection Agency (EPA) Office of Research and Development's
Sustainable and Healthy Communities Research Program (EPA 2015) developed the Human
Well-being Index (HWBI) as an integrative measure of economic, social, and environmental
contributions to well-being. The HWBI is composed of indicators and metrics representing
eight domains of well-being: connection to nature, cultural fulfillment, education, health,
leisure time, living standards, safety and security, and social cohesion. The domains and
indicators in the HWBI were selected to provide a well-being framework that is broadly
applicable to many different populations and communities, and can be customized using
community-specific metrics.
A primary purpose of this report is to adapt the HWBI to quantify human well-being for Puerto
Rico. We compare well-being across Puerto Rican municipios (county-equivalent) and to the
United States, and reflect on the contribution of different indicators and domains to measured
well-being.
Additionally, our adaptation of the HWBI for Puerto Rico provides an example of how the HWBI
can be adapted to different communities. A key technical challenge in doing so is the large
amount of spatial and temporal data that needs to be processed in order to generate
community-specific metrics. Therefore, we provide detailed technical guidance on how spatial
and temporal data can be processed to calculate a community specific HWBI using the freely
available R software (version 3.1.2). Examples from Puerto Rico are used to demonstrate how
the HWBI can be calculated temporally, across multiple spatial scales, and compared across
different communities when data may be community-specific. Puerto Rico presents challenges
in terms of data availability and quality, and recommendations are offered on how to deal with
such challenges. Because we were interested in comparing Puerto Rico to the U.S. States, we
prioritized using the same metrics as the U.S. HWBI whenever possible. The flexibility of the
HWBI framework allows that if different local data is desired or becomes available, it can easily
be substituted in while maintaining the overall integrity of the framework. The key is to make
sure data is consistent across whatever is being compared (e.g., scenarios, zip codes, counties,
years).
Indices, such as HWBI, can help communities to assess and track well-being, and identify
decisions and interventions that can contribute to and promote sustainable well-being. This
research contributes to EPA's Sustainable and Healthy Communities Research Program (EPA
2015) goal of developing research, data, and tools to expand the capabilities of communities to
consider the social, economic, and environmental impacts of decision alternatives on
community well-being.
xii
-------
Chapter 1: Introduction
1.1. Background
The U.S. Environmental Protection Agency (EPA) Office of Research and Development's
Sustainable and Healthy Communities Research Program (EPA 2015) developed the Human
Well-being Index (HWBI) as an integrative measure of economic, social, and environmental
contributions to well-being. The HWBI is responsive to the flows of goods and services within
the natural and built environments (Figure 1.1; Summers et al., 2014; Summers et al., 2016)
and can serve as an endpoint of human well-being within formal decision making frameworks.
For example, publics works projects and emergency preparedness activities (i.e., Social
Services) implemented by city planners to reduce vulnerability to flooding and improve flood
disaster recovery may lead benefits to human well-being through improved safety and security,
human health, living standards, and social cohesion (Summers et al., 2016). Tracked over time,
the HWBI can be used as an indicator of sustainability enabling longer-term evaluation of the
success of alternative resource management and policy decisions (Smith et al., 2014b; Summers
et al., 2014).
The HWBI was based on an intensive literature review of existing global, national, and local
indices of human well-being that comprise varying degrees of economic, environmental, and
socal influences. The domains and indicators selected to describe well-being in the U.S. are
considered broadly applicable to many different populations and communities (Table 1.1).
Indicators and metrics are chosen to describe aspects of human condition, and may include
both subjective and objective data. Correspondingly, metrics describing indicators of well-
being should distinguish measures of human condition (e.g., an emotional connection to
nature; Figure 1.1: Domains of Well-being) from measures of goods and services that can be
modified through decision-making (e.g., access to greenspace; Figure 1.1: Goods and Services)
in order to affect human condition.
The HWBI is unique in that it can be applied to multiple spatial scales, across time, and within
cultural contexts different from the original index (Smith et al., 2013b; Smith et al., 2014a;
Smith et al., 2014b; Summers et al., 2012; United States Environmental Protection Agency,
2012). This has been previously demonstrated for the American Indian Alaska Native population
of the U.S. (Smith et al., 2014b).
Fine-scale adaptations of the HWBI reflect the applicability of the HWBI to a community or
population within the national index. Depending on data availability, metric values can be
based on raw data collected for a specific population or imputed from broader (e.g., national,
state, regional) to finer spatial scales (e.g., county, neighborhood, socioeconomic group). In
addition to scalability, the HWBI can be customized using community-specific measures of well-
being that better reflect the population. The flexibility of individual metrics to suit the cultural
context and data availability of any community extends the transferability of HWBI to
populations outside of the U.S.
1
-------
Figure 1.1. Conceptual diagram linking economic, social, and environmental flows of goods
and services to the HWBI (Summers et al., 2014).
Quality and Quantity of Capital
Good Governance
Ecosystem
-Air Quality
-Food and Fiber
Provisioning
-Greenspace
- Natural Hazard
Protection
-Water Quality
- Water Quantity
Goods and Services
Activism
Communication
Community and
Faith-Based Initiatives
Education
(including inequities)
Emergency
Preparedness
Family Services
Healthcare
(including inequities)
Justice
(including inequities)
Labor
Public Works
Economic
Capital Investment
Consumption
Employment
Finance
Innovation
Production
Re-distribution
(including inequities)
Freedom of Choiceand
Opportunity
Domains of Well-being
Connection to Nature
Leisure Time
Cultural Fulfillment
Living Standards
Education
Safety and Security
Health
Social Cohesion
Well-being Elements
Environmental
Societal
Economic
Human Well-being Index
2
-------
Table 1.1. Descriptions of domains in the Human Well-being Index (from Smith et al. 2012).
Human Well-being Domain
Description
Connection to Nature
Cultural Fulfillment
Education
Health
Measures the emotional affiliation of human beings to
other living organisms and the natural environment
(i.e. biophilia)
Measures opportunities that afford people and
communities access to fulfilling their cultural and
spiritual needs
Measures outcomes derived from the formal and
informal transfer of knowledge and skills, including
basic knowledge and skills, educational attainment and
participation, and various social, emotional, and
development aspects in childhood
Measures a population's health status including
personal well-being, life expectancy and mortality, and
physical and mental health conditions, as well as
lifestyle behavior and healthcare
%
k
9
Leisure Time
Measures time that individuals have to voluntarily
engage in pleasurable activities when they are free
from the demands of work or other responsibilities
Living Standards
Measures the physical circumstances in which people
live, including access to economic resources and
attainment of basic human needs (e.g., food, shelter)
Safety and Security
Social Cohesion
Measures actual and perceived freedom from harm
including physical safety, national security, and
financial security
Measures the ties that bind humans together in society
through healthy and active social networks that allow
open discussion and resolution of problems, and give
members a sense of identity
3
-------
1.2. Adapting the HWB1 to Puerto Rico
We adapted the HWBI to compare human well-being both within Puerto Rico and to the U.S
population. Puerto Rico presents an opportunity to further explore the transferability of the
HWBI to a population that is linked to the U.S. government, economy, and institutions, yet
culturally distinct and geographically isolated from the mainland U.S. (Figure 1.2). Because
Puerto Rico is a territory of the United States with commonwealth status, it operates
simultaneously as a state-equivalent and as an independent entity. Similar to a U.S. state,
Puerto Rico is governed by both the national and locally elected governments, residents of
Puerto Rico have U.S. citizenship, and the economy is within the U.S. banking and financial
system (Bram et al., 2008; Central Intelligence Agency, 2015).
Figure 1.2. Map of Puerto Rico in relation to contiguous United States (source: Esri).
In 2014, Puerto Rico's population was estimated at 3,5 million residents, comparable to the 29th
most populated U.S. state (United States Census Bureau, 2014a). Similar in size to the U.S. state
of Connecticut (Bram et al., 2008), it is densely populated with the majority of residents (2.1
million) in metropolitan San Juan, the 21st most populated urban area in the 2010 U.S. Census
4
-------
(United States Census Bureau, 2012a). Dense population combined with geographic isolation
and institutional differences have presented unique challenges for Puerto Rico. During the mid-
20th century, the economy of Puerto Rico transitioned from agriculture to industry, resulting in
many rural residents without work and contributing to a mass migration of 1 million residents
to the mainland U.S. during 1950 - 2000 (Bram et al., 2008). Although investments in
specialized industries like pharmaceuticals prospered and the Puerto Rico economy was
stronger than most of the Caribbean, economic recession after 2005 furthered the migration to
the mainland and has caused a net population loss on the island in recent years that ranks
among the highest globally (Abel and Deitz, 2014; Central Intelligence Agency, 2015).
Adapting the HWBI to Puerto Rico builds on the existing framework and allows for Puerto Rico
to be included in local- and national-level decision and policy making. We considered data
availability within the context of Puerto Rico communities and in relation to the U.S. in order to
measure well-being distinctive to Puerto Rico while maintaining the integrity of the HWBI for
comparisons with the U.S. For this adaptation, we scaled the HWBI to the 78 municipios (Figure
1.3) which make up Puerto Rico and function as county-equivalents according to the U.S.
Census Bureau.
Desecheo
Island
Mona Island
Aguadilia Isabels
Quebradillas :
¦ Camuftatnio ™
Moca.
Aguada ./ :
Rincta San Sebastian
Anasco ;,
"> .-••• Lares.
Las Marias :
Utuado
Mayagiiez
Maricao Adjuntas
Hormigueros , i ; Villalba;
San German Yauco " Cosmo
Sabana Grande penuelas «
renueifs p0nc® juana Dfaz.-
Guayarillla •
Barcelona Vega Baja Dorall R3Eatafio
j Manati VegaAIia 1 San Juan'
Florida Toa Alta . . Carolina ^ Luquillo
Bayamon TruiiHoAlto ftio Grande
,1 . Corozar""'"*-.; G«aynab0 ; ; C,n6v,nas • Fajardo
Morovis ' - '• \ ¦¦
.•¦.Ciates ! Naranj.lo .
-. Agyas Buenas ura ° ' Cesba
\ „ . r Com«to-t Juncos' NaguaUo
Javuva •• Orocovis . Caguas .
y y . ; Barranquitas Las Piedras
: San Lorenzo Humacao
Aibonlto
Cayey
Yabucoa
Cabo Rojo
Lajas
Patillas
Guanica
Santa Isabel
rdtaiicis
Salinas Guayama Maunabo
Arroyo
Vieques
Figure 1.3. Map of Puerto Rico including boundaries of 78 municipios. Outlying islands of
northwestern Puerto Rico (Mona Island and Desecheo Island) are uninhabited ecological
reserves included within Mayaguez municipio. For this figure, outlying islands were shifted
closer to mainland Puerto Rico (source: US Census).
Puerto Rico has many assets and needs that are comparable to the U.S. but also has disparities
that challenge decision makers. For example the Puerto Rico economy is publicized as
considerably less affluent than the mainland U.S. (Bram et al., 2008) with high public debt and
credit default (Central Intelligence Agency, 2015). In 2010, median household income was 62%
lower in Puerto Rico than the mainland U.S. (United States Census Bureau, 2015a) and Puerto
5
-------
Ricans on the mainland during the post-recession period 2010-2012 fared considerably better
than those living on the island (United States Census Bureau, 2014b). Another highly publicized
disparity is safety and security. In 2010, the murder and non-negligent manslaughter rate in
Puerto Rico exceeded the highest in the mainland U.S., but has shown declines in recent years
(New Orleans-Metairie-Kenner Louisiana Metropolitan Statistical Area; FBI, 2015). Puerto
Ricans are also disproportionately affected by asthma, with child asthma prevalence as much as
2 times higher in Puerto Rico than in the U.S. states (CDC 2011).
Yet Puerto Rico has higher home ownership compared to the mainland U.S. and substantially
fewer homes with mortgages (United States Census Bureau, 2015a). Furthermore, Puerto Rico
had among the highest rate of respondents that considered themselves very happy compared
globally and to the U.S. (World Values Survey, 2015). College education rates are similar
between Puerto Rico and the mainland U.S., although high school graduation rates are much
lower (United States Census Bureau, 2015a). Although cancer was the leading cause of illness
and death in Puerto Rico for the period 2007-2011, compared to the mainland U.S., Puerto Rico
has a lower overall age-adjusted incidence rate of cancer than the U.S. and the lowest incidence
rate of new lung cancer, attributed to having among the lowest smoking rates (CDC, 2015c).
6
-------
Chapter 2: Identifying Metrics and Finding Data
2.1. ucture
The HWBI endpoint is a multidimensional composite value interpreting human well-being.
Although this is a holistic index of overall well-being, the HWBI can be further separated into
sub-indices of economic, social, or environmental well-being (Summers et al., 2014). The index
is made up of eight domains of well-being—Connection to Nature, Cultural Fulfillment,
Education, Health, Leisure Time, Living Standards, Safety and Security, and Social Cohesion-
considered globally applicable to human well-being (Smith et al., 2013b; Summers et al., 2014).
Domains are organized into 25 indicators comprised of 80 individual metrics measuring various
components of economic, social, and environmental well-being (Figure 2.1). The HWBI integrity
is maintained at the indicator-level; within the scope of an indicator, metrics are flexible and
can be modified to suit data availability or community context (Smith et al., 2014a).
¦ If, 1" hi k'" /!• •ctic-i i i'm [ i "erto Rico HWBI
As a commonwealth territory of the U.S., Puerto Rico is simultaneously administered at the
national- commonwealth- (state equivalent), and municipio-level (county equivalent). As such,
metric data were collected from global (e.g., World Values Survey), national (e.g., U.S. Census
Bureau), and commonwealth (e.g., Puerto Rico Department of Education) sources. Metric
measures and sources were prioritized for those selected for the U.S. HWBI, based primarily on
publically available data from a reputable agency collected with spatial and temporal
continuity. A comparative summary of each metric for Puerto Rico and the U.S. HWBI is
included in Appendix A.
The U.S. HWBI was calculated from 80 metrics (Table 2.1). Of these 80, the Puerto Rico HWBI
had 30 metrics available that were the same as their corresponding U.S. metric and from the
same U.S. source, 24 metrics the same U.S. measure but from a different source, one metric as
an alternate to the U.S. measure, 16 metrics substituted with different measures but with the
same units as the U.S. version, three metrics substituted with different units than the original
U.S. metrics, and six not available with no substitutes identified (Table 2.2). Within HWBI,
domains and indicators vary in terms of number of metrics. Of the indicators, three were of
concern during metric selection (Time Spent, Perceived Safety, and Risk) because the indicators
were comprised of only a single metric and therefore Puerto Rico metric data were necessary
to maintain the integrity of the index.
7
-------
Figure 2.1. Diagrammatic representation of human well-being components. HWBI is built
from 80 metrics and organized into 25 indicators describing 8 domains of well-being considered
globally applicable to human well-being. Metric details are described in Appendix A.
ALLOFLFE
BEAUSPRT
Participation
Actual Safety
\Attainment
Basic
Educational
Knowledge &
Skills of
Youth
Biophilia
Social,
Emotional &
)evelopmental
Aspects
Perceived
Safety
ASTHMORT
CANCMORT
DIABMORT
HRTDISMORT
INFMORT
LIFEXPCT
SUICDMORT
Connection
to Nature
Safety &
Security
POLINTRST
REGVOTRS
SATDEM
TRUSTGOV
VOICENGOV
VOTRTOUT
Healthcare
Education
Life
Expectancy &
Mortality
Social
Engagement
Democratic
Engagement
ALCOHOL
HBI
TEENPREG
TEENSMK
Lifestyle &
Behavior
CHLDREAD
MEALS
WATCHTV
Social
Cohesion
Family Bonding
Human Well-being
Health
Physical&
Mental Health
Conditions
ADLTASTHMA
CANCER
CHLDASTHMA
DEPRESSION
DIABETES
HRTATTACK
HRTDISEASE
OBESITY
STROKE
Attitude^^
toward Others &
Community r
CANTRUST
CITYSATIS
CLSETOWN
DISCRM2
HELPFUL
Personal
Well-being
Social
Support
Living
Standards
Cultural
Fulfillment
Leisure
Time
ultural Activit
Participation
Wealth
eisure Activit
Participation
Basic
Necessities
Working Age
Adults
Domain
Income
Work
Indicator
LEISURE
METRIC
ALLOFLFE: Connection to all of life
ACCMM: Accidental death rate
ADLTASTHMA: Adult asthma rate
ADULTLIT: Adult literacy rate
ALCOHOL: Alcohol consumption
ASTHMORT: Asthma mortality rate
BEAUSPRT: Spirituality with nature
BULLY: Children feeling unsafe
CANCER: Adult cancer rate
CANCMORT: Cancer mortality rate
CANTRUST: Trust in people
CHLDACTV: Child activity participation
CHLDASTHMA: Child asthma rate
CHLDHLTH: Children's health
CHLDREAD: Time reading to children
CHLDSOCIAL: Child social behavior
CITYSATIS: City satisfaction
CLSETOWN: Feeling toward one's town
CLSFRNDFAM: Close family or friends
CON FACT: Parents read to children
DEPRESSION: Adult depression rate
DIABETES: Adult diabetes rate
DIABMORT: Diabetes mortality rate
DISCRM2: Racial discrimination
EMTSUPRT: Getting emotional support
FAMDOC: Regular doctor visits
FOODSECURE: Food security
GRPACTV: Organized participation
HAPPY: Happiness
HBI: Healthy Behaviors Index
HELPFUL: Perceived helpfulness
HOMEAFFORD: Home affordability
HOMEVAL: Median home value
HRTATTACK: Heart attack rate
HRTDISEASE: Heart disease rate
HRTDISMORT: Heart disease mortality
HSGRAD: High school graduation
INFMORT: Infant mortality rate
JOBLOSE: Fear of job loss
JOBSATIS: Job satisfaction
LEISURE: Time spent relaxing
LIFESATIS: Life satisfaction
LIFEXPCT: Life expectancy at birth
LONGWRKHRS: Long work hours
MATHTEST: Standardized math test
MEALS: Family meal time
MEDINCOME: Median income
MTGDEBT: Mortgage debt
NATHAZHLOSS: Natural hazard injury
NORMWRKHRS: Daytime work hours
OBESITY: Adult obesity rate
PARTNEDU: College enrollment
PERARTS: Attend performing arts
PHYSACTIV: Physical activity
POLINTRST: Interest in politics
POVERTY: Poverty rate
POVPERSIST: Persistent poverty rate
PRCVDHLTH: Perceived health
PRCVDSAFE: Perceived safety
PROPCRIME: Property crime rate
READTEST: Standardized reading test
REGVOTRS: Registered voters
SATDEM: Satisfaction with democracy
SATISHLTHC: Hospital care satisfaction
SCITEST: Standardized science test
SENIORCARE: Adult care activities
SOVI: Social Vulnerability Index
STROKE: Lifetime adult stroke rate
SUICDMORT: Suicide mortality rate
TEENPREG: Teen pregnancy rate
TEENSMK: Teen smoking rate
TOTRATE: Religious affiliation
TRUSTGOV: Trust in government
UNIVGRAD: College graduation rate
VACATION: Time spent on vacation
VIOLCRIME: Violent crime rate
VOICENGOV: Voice in government
VOLNTR: Volunteer rate
VOTRTOUT: Voter turnout
WATCHTV: Time watching television
8
-------
Table 2.1. Summary of 80 metrics composing 25 indicators within 8 domains of well-being.
See Appendix A for detailed metric descriptions.
Domain
Indicator
Metric(s)
Metric Description
Connection to
Biophilia
ALLOFLFE
Connection to all of life
Nature
BEAUSPRTC
Spiritually touched by beauty of creation
Cultural
Cultural Activity
PERARTSb
Performing arts attendance
Fulfillment
Participation
TOTRATE
Belonging to religious denomination
Education
Basic Educational
MATHTEST
Standardized math test achievement
Knowledge and Skills of
READIEST
Standardized reading test achievement
Youth
SCITEST
Standardized science test achievement
Participation and
ADULTLIT
Adult literacy rate
Attainment
HSGRAD
High school graduation rate
PARTNEDU
Post-secondary education enrollment
UNIVGRAD
Post-secondary education graduation rate
Social, Emotional, and
BULLY
Children feeling unsafe at school
Developmental Aspects
CHLDHLTH
Children's health
CHLDSOCIAL
Children's social behavior
CON FACT0
Parents who read to children
Health
Healthcare
FAMDOC
Regular doctor visits
SATISHLTHC
Satisfaction with hospital care
Life Expectancy and
ASTHMORT
Asthma mortality rate
Mortality
CANCMORT
Cancer mortality rate
DIABMORT
Diabetes mortality rate
HRTDISMORT
Heart disease mortality rate
INFMORT
Infant mortality rate
LIFEXPCT
Life expectancy at birth
SUICDMORT
Suicide mortality rate
Lifestyle and Behavior
ALCOHOL
Alcoholic beverage consumption
HBI
Healthy Behaviors Index
TEENPREG
Teen pregnancy rate
TEENSMK
Teen smoking rate
Personal Well-being
HAPPY
Happiness
LIFESATIS
Life satisfaction
PRCVDHLTH
Perceived health
Physical and Mental Health
ADLTASTHMA
Lifetime adult asthma rate
Conditions
CANCER
Lifetime adult cancer rate
CHLDASTHMA
Lifetime child asthma rate
DEPRESSION
Lifetime adult depression rate
DIABETES
Lifetime adult diabetes rate
HRTATTACK
Lifetime adult heart attack rate
HRTDISEASE
Lifetime adult heart disease rate
OBESITY
Adult obesity rate
STROKE
Lifetime adult stroke rate
9
-------
Table 2.1. continued.
Domain
Indicator
Metric(s)
Metric Description
Leisure Time
Leisure Activity Participation
PHYSACTIV
Physical activity participation
VACATION"
Time spent on vacation
Time Spent
LEISURE3
Time spent on leisure or relaxing
Working Age Adults
LONGWRKHRS
Long work hours
NORMWRKHRS
Regular daytime work hours
SENIORCAREc
Adult care activities
Living
Basic Necessities
FOODSECURE
Food security
Standards
HOMEAFFORD
Home affordability
Income
MEDINCOME
Median household income
POVERTY
Poverty rate
POVPERSIST
Persistant poverty rate
Wealth
HOMEVAL
Median home value
MTGDEBT
Mortgage debt
Work
JOBLOSE
Fear of job loss
JOBSATIS
Job satisfaction
Safety and
Actual Safety
ACCMM
Accidental death rate
Security
NATHAZHLOSS
Natural event injury and death rate
PROPCRIME
Property crime rate
VIOLCRIME
Violent crime rate
Perceived Safety
PRCVDSAFE
Perceived safety
Risk
SOVIa
Social Vulnerability Index
Social Cohesion
Attitude toward Others and
CANTRUST
Trust in people
the Community
CITYSATIS
City satisfaction
CLSETOWN
Feeling close to one's town or city
DISCRM2C
Racial discrimination
HELPFUL
Perception that others are helpful
Democratic Engagement
POLINTRST
Interest in politics
REGVOTRS
Registered voters
SATDEM
Satisfaction with democracy
TRUSTGOV
Trust in government
VOICENGOV
Voice in government
VOTRTOUT
Voter turnout
Family Bonding
CHLDREADC
Time spent reading to children
MEALSC
Family meal time
WATCHTV
Time spent watching television
Social Engagement
CHLDACTV
Child organized activity participation
GRPACTV
Participation in organized group
VOLNTR
Volunteer rate
Social Support
CLSFRNDFAM
Having close family or friends
EMTSUPRT
Getting emotional or social support
aMetrics were imputed for comparisons to U.S. HWBI; removal would result in loss of indicator
bMetrics were removed for comparisons to U.S. HWBI
cMetrics were not available and were not imputed for comparisons to U.S. HWBI
10
-------
Table 2.2. Summary of metric availability for Puerto Rico compared to U.S. HWBI. See Figure
2.2 for definitions of metric availability categories and Appendix A for metric descriptions.
Same as
Modified
Alternative
Substitute
Substitute measure
No data
U.S. HWBI
source
U.S. HWBI
measure
and units
available
ACCMM
ADULTLIT
HBI
ALLOFLFE
PERARTSb
BEAUSPRTC
ADLTASTHMA
CANTRUST
CHLDACTV
SOVIa
CHLDREADC
ALCOHOL
GRPACTV
CHLDHLTH
VACATIONb
CONFACTc
ASTHMORT
HAPPY
CHLDSOCIAL
DISCRM2C
BULLY
LIFEXPCT
CITYSATIS
MEALSC
CANCER
LONGWRKHRS
CLSETOWN
SENIORCAREc
CANCMORT
MATHTEST
CLSFRNDFAM
CHLDASTHMA
MEDINCOME
FOODSECURE
DEPRESSION
NATHAZHLOSS
HELPFUL
DIABETES
NORMWRKHRS
JOBLOSE
DIABMORT
OBESITY
JOBSATIS
EMTSUPRT
PARTNEDU
LEISURE3
FAMDOC
POLINTRST
POVPERSIST
HOMEAFFORD
POVERTY
SATISHLTHC
HOMEVAL
PRCVDSAFE
TRUSTGOV
HRTATTACK
PROPCRIME
VOICENGOV
HRTDISEASE
READTEST
HRTDISMORT
REGVOTRS
HSGRAD
SATDEM
INFMORT
SCITEST
LIFESATIS
TOTRATE
MTGDEBT
VIOLCRIME
PHYSACTIV
VOLNTR
PRCVDHLTH
VOTRTOUT
STROKE
SUICDMORT
TEENPREG
TEENSMK
UNIVGRAD
WATCHTV
aMetrics were imputed for comparisons to U.S. HWBI; removal would result in loss of indicator
bMetrics were removed for comparisons to U.S. HWBI
cMetrics were not available and were not imputed for comparisons to U.S. HWBI
11
-------
lute Metrics
Although many agencies collect data for Puerto Rico as a state-equivalent, there are many
instances where data were not available and we relied on previously demonstrated methods of
metric substitution or imputation. Previously, the HWBI framework was scaled to the American
Indian Alaska Native population of the U.S. and metric availability at the population-specific
scale and cultural relevance were assessed. To maintain the integrity of the HWBI, metric
alternates were selected that matched both the representation of the substance of the
measure and type of data (e.g., quantitative or qualitative) (Smith et al., 2014b).
For Puerto Rico, metrics were selected based on data availability and relevance to Puerto Rico,
prioritizing U.S. HWBI metric sources while making necessary modifications of metric sources
and measures to maintain the integrity and comparability of the index (Figure 2.2). Substitute
metric measures were identified that were either similar with the same units for direct
substitution and comparison to the U.S. HWBI (U.S. interquartile ranges were used for
normalization and standardization) or similar but with different units and therefore not
comparable to the U.S. (Table 2.2).
We followed the conceptual framework of U.S. HWBI's adaptation to native populations (Smith
et al., 2014b) and the Organisation for Economic Co-operation and Development's Better Life
Index (Organisation for Economic Co-operation and Development, 2011), which faced the
challenge of compiling and comparing data for numerous countries. Some substitute metrics
were based on non-governmental data and may not have been collected with the same sample
size and estimation standards. However, these substitute measures and sources were better
than removing a metric or imputing from outside of Puerto Rico. Substitutions may also
elucidate gaps in data availability, such as connection to nature and cultural fulfillment, which
could be improved by local data collection agencies.
Because we were interested in comparing Puerto Rico to the U.S. states, we prioritized using
the same metrics as the U.S. HWBI whenever possible, rather than allowing open flexibility in
choosing metrics that may have greater local relevance for Puerto Rico. Unique attributes of
Puerto Rican culture (e.g., traditional cooking, informal neighborhood gatherings) or
underground economy (e.g., street vendors, roadside stands, cash only businesses) may not be
reflected in national-scale statistics. The flexibility of the HWBI framework allows that if
different local data is desired or becomes available, it can easily be substituted in while
maintaining the overall integrity of the framework. The key is to make sure data is consistent
and at a meaningful scale for whatever is being compared (e.g., zip codes, counties, years).
12
-------
Figure 2.2, Decision tree used to select metric data for Puerto Rico adaptation of HWBI.
No-
Yes
Yes
No-
Yes
No-
Yes-
No-
Yes
No-
Modified source
Alternate U.S. HWBI
Same as U.S. HWBI
No data available
Are data available for
the same measure and
same source as U.S.
HWBI?
Are data available for a
different but similar
measure with the same
units as U.S. HWBI?
Substitute measure
and units
Substitute measure
Are data available for a
different but similar
measure with different
units than U.S. HWBI?
Are data available for
the alternate measure
from the alternate
source of U.S. HWBI?
Are data available for
the same measure but a
different source than
U.S. HWBI?
13
-------
2.4. Spatial Availability of Metric Data
Metric data differed in spatial availability. These scales included availability by Commonwealth
of Puerto Rico (state equivalent), municipio (county equivalent), Public Use Microdata Area
(PUMA) (U.S. Census Bureau's statistical areas based on population), Metropolitan Statistical
Area (MSA) (U.S. Office of Management and Budget's statistical areas based on urbanization),
and World Values Survey region (Figure 2.3).
Figure 2.3, Maps of Puerto Rico with boundaries for HWBI metric data spatial scales including
by a) Commonwealth of Puerto Rico (state equivalent), b) municipio (county equivalent), c)
Public Use Microdata Area (PUMA) based on 2000 U.S. Census, d) PUMA based on 2010 U.S.
Census, e) Metropolitan Statistical Area (MSA) based on 2000 U.S. Census, f) MSA based on
2010 U.S. Census, and g) World Values Survey region. For this figure, outlying islands were
shifted closer to mainland Puerto Rico.
14
-------
Further complicating data analysis with multiple scales is the consideration that spatial scales
can overlap and boundaries change over time. For example, U.S. census boundaries change
according to decennial census population. Geographic information systems (GIS) analyses in
ArcMap 10.1-10.3 were used to overlay boundaries and assign each municipio x year
combination to a coarser spatial unit. For Puerto Rico PUMA-level data, boundary delineations
based on the 2000 U.S. Census were used for 2000-2011 and those of the 2010 U.S. Census for
2012-2013 (following the reporting by the U.S. Census Bureau). For MSA-level data, boundary
delineations from 2000 were used for 2000-2002, from 2003 for 2003-2011, and from 2012 for
2012-2013 (following the reporting by the U.S. Office of Management and Budget). As
aggregate boundaries do not necessarily follow county boundaries (e.g., a single county can be
in multiple PUMAs) overlay analyses were also used to assign each municipio to a single
aggregate area by year, according to majority land area.
15
-------
Chapter 3: Downloading and Organizing Data
i I . ill « diii'- 1! 1 I>1 LVta
The data for Puerto Rico was downloaded for all available metrics (Table 2.1, Table 2.2). Data
were collected from numerous sources (Appendix A) through extraction from numerous
databases available in a variety of formats. As such, the collection of downloaded metric data
required organization to ensure compatibility among metrics and for HWBI processing (Chapter
4) and calculation (Chapter 5).
Publically available, web-based governmental or non-governmental organization databases
were used to extract most of the data in a summarized form (i.e., raw data were already
tabulated as a percentage or rate for a specific area). Additionally, data were derived from
peer-reviewed literature, reports, or other publications (Appendix A). It was assumed that all
data were summarized according to appropriate statistical practices in terms of sample size,
randomization, and confidence limits. If a metric value was reported with a geographic
identifier (e.g., county), it was assumed that appropriate considerations were made by the
source to ensure it was representative of an area. In some cases, raw occurrence data were the
only data available (e.g., individual mortality events) and we summarized the data for an area.
Following the Centers for Disease Control and Prevention's Behavioral Risk Factor Surveillance
System protocol, we considered these summaries reliable if they were based on at least 50
observations. Otherwise data were collapsed to a coarser spatial scale. Additionally for metrics
based on occurrence data, a finer spatial scale (e.g., a large county with data) could be
subtracted from a coarser spatial scale (e.g., state, MSA) so that the resulting value of the
coarser spatial scale was independent of the finer spatial scale data (e.g., state value without
large counties). Although this could conceptually extend to most metrics by gathering
occurrence data instead of pre-summarized data, web-based database extraction of spatially
and temporally summarized data was favored over raw data where available.
Most data were modified from the original downloaded format within R to maintain the
original variable names and file structure for reference while allowing for consistent formatting
among metrics. During this download and organization process, individual datasets were
condensed into a single dataset with all metric x county x year combinations. Alternative data
management approaches could improve efficiency or flexibility by incorporating database
software (e.g., Microsoft Access) or automating data downloads and organization through an
application program interface (API) in combination with programming.
16
-------
* * Spah II s.ii M • iii| « i II Iiu ieoff
There is a trade-off between spatial scale and time. For example, many government agencies
(e.g., U.S. Census Bureau, Centers for Disease Control and Prevention) only reported data by
county if there was a sufficient sample size to develop a reliable estimate for the population
and sufficient population size as not to identify a single individual from a dataset. Otherwise
annual data are not reported by county, but aggregated as a PUMA or MSA. Alternatively,
county-level data could be aggregated over time (3- or 5-year estimates) to increase sample
size and allow for reporting. For Puerto Rico HWBI, annual metric data were favored for
available municipios and data aggregated by PUMA or MSA were used for remaining
municipios. Time-aggregated data were used in cases where county-level data were necessary
to calculate a metric (e.g., age-specific population data were needed to calculate a rate or
percentage).
* - > »i, . iii ill' . h i Formatting Downloaded Metii. tvta
Data were downloaded and organized into folders by metric, source, spatial scale, and/or year.
For example, the median value of owner-occupied housing units (HOMEVAL) data were
downloaded, organized by metric. Although HWBI data were based on county-level geography,
not all HOMEVAL data were available at that scale and were additionally organized by
geographic availability (e.g., "HWBI\Data_Downloads\HOMEVAL\PR_Municipio"). Similarly,
datasets containing multiple metric data were organized by source. For example, Multiple
Cause of Death data from Centers for Disease Control and Prevention were organized by source
(e.g., "HWBI\Data_Downloads\CDC_Mortality").
Here we present three examples of organizing and formatting metric data in R version 3.1.2
(http://www.r-project.org) using 'tidyr' version 0.2.0 (Wickham, 2015b) and 'dplyr' version
0.4.1 (Wickham, 2015a). These include examples of managing data based on multiple spatial
scales, using multiple variables to calculate metric values, and summarizing data based on
occurrence data.
Example 1: Organizing and formatting metric data based on multiple spatial scales
Many HWBI metrics were available for multiple spatial scales, with the finest available for more
populated municipios. HOMEVAL data were downloaded from the U.S. Census Bureau,
American Community Survey using the web-based database American FactFinder (Figure 3.1;
United States Census Bureau, 2015a). These data were available for 2005-2013 by county,
PUMA, and/or state.
17
-------
U.S. Department of Commerce
AMERICAN
United States'
Census
FactFi
nder I >
Bureau
MAIN COMMUNITY FACTS
GUIDED SEARCH
ADVANCED SEARCH DOWNLOAD CENTER
Search - Use the options on the left (topics, geographies,...) to narrow your search results
Your Selections
Search using...
Search:
B25077 O
Dataset:
2012 ACS 1-year estimates O
County
Al Counties within Puerto Rico 0
clear al selections and
start a new search
load search | save search
Search using the options below:
Topics
(age, income, year, dataset....)
Geographies
(states, counties, places....)
Race and Ethnic Groups
(race, ancestry, tribe)
Industry Codes
(NAICS industry....)
EEO Occupation Codes
(executives, analysts....)
Search Results: 1-1 of 1 tables and other products match "Your Selections'
Refine your search results:
topic or table name
state, county or place (optional)
D l l El ©
= • topics race/ancestry industries occupations
1 Selected:
1
View [£] Download
Compare [r] Clear All
$ Reset Sort
©
< 1
~
Show results from: | 2012 (0)
~
| Al available programs
T
ID
Table, File or Document Title
Dataset £
About
£) B2S077
MEDIAN VALUE (DOLLARS)
2012 ACS 1-year estimates
©
1 Selected:
View j£] Download
ik] Compare [r] Clear All
$ Reset Sort
©
44
< 1
~
»
Figure 3.1. Sample screen-shot of U.S. Census Bureau's web based interface for downloading
U.S. Census Data (www.factfinder.census.gov). In this example, median home value data were
selected by year (2012) and county for Puerto Rico.
In this example, we present the unprocessed 2012 median home value data by municipio
(Table 3.1) and formatting for the HWBI (Table 3.2). Since county-level data were only available
for more populated municipios, we also downloaded (Table 3.3) and formatted (Table 3.4) data
aggregated by PUMA. These multiple spatial scales result in overlapping data for municipios
with both county-level and regional availability (e.g., Arecibo municipio) (Table 3.4). Later in
this chapter, we demonstrate the selection of data for only the finest available spatial scale for
each metric x county x year combination.
18
-------
Box 3.1. R Code to read in municipio-level HOMEVAL data for 2012.
# Read in municipio-level HOMEVAL data for 2012
xl2 <- read. csv("./HWBI/Data_Downloads/HOMEVAL/PR_Municipio/ACS_12_lYR_B25077_with_ann.csv",
header=T, stringsAsFactors=F)
xl2$Year <-"2012"
Table 3.1. Unformatted and unprocessed 2012 median home value data (U.S. dollars;
HD01_VD01) by county for available municipios (GEO.id2) downloaded from U.S. Census
Bureau, American Community Survey using the web-based database American FactFinder.
GEO.id
GEO.id2
GEO.display.label
HD01VD01
Year
0500000US72013
72013
Arecibo Municipio, Puerto Rico
107100
2012
0500000US72021
72021
Bayamon Municipio, Puerto Rico
144400
2012
0500000US72025
72025
Caguas Municipio, Puerto Rico
143300
2012
0500000US72031
72031
Carolina Municipio, Puerto Rico
152700
2012
0500000US72061
72061
Guaynabo Municipio, Puerto Rico
212400
2012
0500000US72097
72097
Mayaguez Municipio, Puerto Rico
102100
2012
0500000US72113
72113
Ponce Municipio, Puerto Rico
109800
2012
0500000US72127
72127
San Juan Municipio, Puerto Rico
163800
2012
0500000US72135
72135
Toa Alta Municipio, Puerto Rico
163500
2012
0500000US72137
72137
Toa Baja Municipio, Puerto Rico
135600
2012
0500000US72139
72139
Trujillo Alto Municipio, Puerto Rico
163100
2012
Box 3.2. R Code to format county-level metric data for 2012 median home value.
x<-xl2
x <- x[,c("GEO.id2","HD01_VD01","Year")]
colnames(x)[colnames(x)=="HD01_VD01"] <- "HOMEVAL_m" # rename census variable with HWBI variable name
x$GEO.id2 <- as.character(x$GEO.id2)
HWBI <- left_join( municipio, x, by= c("CNTYIDFP"="GEO.id2", "Year")) #join data to county x year template
Table 3.2. Sample of formatted county-level metric data for 2012 median home value (U.S.
dollars) by municipio for available municipios (HOMEVAL_m).
NAME
CNTYIDFP
GEO.id2_00
GEO.id2_10
Year
HOMEVALm
Adjuntas
72001
7202200
7200401
2012
NA
Aguada
72003
7200100
7200101
2012
NA
Aguadilla
72005
7200100
7200102
2012
NA
Aguas Buenas
72007
7202400
7200602
2012
NA
Aibonito
72009
7202400
7200602
2012
NA
Arecibo
72013
7200300
7200301
2012
107100
19
-------
Box 3.3. R Code to read in PUMA-level HOMEVAL data for 2012 and apply data to PUMA 2010
geography.
# Read in PUMA-level HOMEVAL data for 2012
xl2 <-
read. csv("L:/Priv/Sustainable_Community_Projects/SanJuan/Projects/HWBI/Data_Downloads/HOMEVAL/PR_PU
MS/ACS_12_lYR_B25077_with_ann.csv", header=T, stringsAsFactors = F)
xl2$Year <-"2012"
xlO <- xl2 # apply data to PUMA 2010 geography
xlO <- xl0[,c("GEO.id2","HD01_VD01","Year")]
colnames(xl0)[colnames(xl0)=="HD01_VD01"] <- "HOMEVAL_plO"
xlO$GEO.id2 <- as.character(xlO$GEO.id2)
HWBI <- left_join(HWBI, xlO, by=c("GEO.id2_10" = "GEO.id2", "Year")) #join to master dataframe
Table 3.3. Unformatted and unprocessed 2012 median home value data (U.S. dollars;
HD01_VD01) by county for available Public Use Microdata Areas (PUMAs; GEO.id2)
downloaded from U.S. Census Bureau, American Community Survey using the web-based
database American FactFinder.
GEO.id
GEO.id2
GEO.display.label
HD01VD01
Year
7950000US7200101
7200101
San Sebastian, Aguada, Moca, Anasco & Rincon
Municipios-Carr 2-Carr 111 PUMA; Puerto Rico
121100
2012
7950000US7200102
7200102
Aguadilla, Isabela & Quebradillas Municipios-Carr 2
(Noroeste) PUMA; Puerto Rico
118600
2012
7950000US7200201
7200201
Cabo Rojo, San German, Lajas & Sabana Grande
Municipios-Carr 2-Carr 100 PUMA; Puerto Rico
106600
2012
7950000US7200202
7200202
Mayaguez, Hormigueros, Las Marias & Maricao
Municipios-Carr 2 (Suroeste) PUMA; Puerto Rico
105900
2012
7950000US7200301
7200301
Arecibo, Barceloneta & Florida Municipios-Carr 2
(Norte) PUMA; Puerto Rico
111300
2012
7950000US7200302
7200302
Hatillo, Camuy, Utuado & Lares Municipios-Carr 2-Carr
129 PUMA; Puerto Rico
121100
2012
Table 3.4. Sample of formatted county-level metric data for 2012 median home value (U.S.
dollars) by municipio for available municipios (HOMEVAL_m) and Public Use Microdata Areas
(PUMAs; HOMEVAL_plO) using 2010 U.S. Census boundaries.
NAME
CNTYIDFP
GEO.id2_00
GEO.id2_10
Year
HOMEVALm
HOMEVALplO
Adjuntas
72001
7202200
7200401
2012
NA
89500
Aguada
72003
7200100
7200101
2012
NA
121100
Aguadilla
72005
7200100
7200102
2012
NA
118600
Aguas Buenas
72007
7202400
7200602
2012
NA
125300
Aibonito
72009
7202400
7200602
2012
NA
125300
Arecibo
72013
7200300
7200301
2012
107100
111300
20
-------
Example 2: Organizing and formatting metric data based on multiple variables
Some HWBI metrics were calculated as a compound value (e.g., rate, percentage) from multiple
input variables. For example, the registered voters (REGVOTRS) metric was calculated as the
percentage of the population (aged 18 and older) registered to vote. Registered voter data was
downloaded from Comision Estatal de Elecciones available for 2004, 2008, and 2012 by voting
district (Comision Estatal de Elecciones, 2015). Population data by age was downloaded from
the U.S. Census Bureau, American Community Survey using the web-based database American
FactFinder available as 5-year estimates (2009-2013) by municipio. In this example, we present
the unprocessed 2012 voter registration data (count of registered voters) by voting district
(Table 3.5). These datasets were combined with population data to calculate the percentage of
the population registered to vote (Table 3.6).
Box 3.4. R Code to read in 2012 data based on senatorial district maps.
# 2012 data based on 2011 senatorial district maps
xl2 <-
read. csv("L:/Priv/Sustainable_Community_Projects/SanJuan/Projects/HWBI/Data_Downloads/VOTRTOUT_REGVO
TRS/Elecciones_en_Puerto_Rico_2012.csv", header=T, stringsAsFactors = F)
Table 3.5. Sample of unformatted and unprocessed 2012 voter registration data (Electores) by
Senate District and Precinct (Precinto) downloaded from Comision Estatal de Elecciones.
Precinto
X
Electores
Votaron
X.
Distrito Senatorial
NA
Numero 1 - San Juan
1
San Juan
54,265
38,718
71.3
2
San Juan
52,766
39,920
75.7
3
San Juan
52,327
38,939
74.4
4
San Juan
58,336
46,770
80.2
5
San Juan
16,175
13,042
80.6
6
Guaynabo
23,652
19,143
80.9
81
Aguas Buenas
19,381
15,572
80.1
Total
276,902
212,104
76.6
21
-------
Box 3.5. R Code to calculate percent of population registered to vote.
ft Election Turnout by Senate Districts and Precincts (includes municipio identifier)
xl2 <- xl2[xl2$X!="", ] ft remove blank rows from .csv
xl2 <- left_join(xl2, lookupM[,c("CNTYIDFP","NAME")], by=c("X"="NAME"))
xl2 <- xl2[,c("CNTYIDFP","Electores","Votaron")]
xl2$Electores <- as.numeric(sub(",",xl2$Electores)) ft remove and convert to numeric
xl2$Votaron <- as. numeric(sub(",",xl2$Votaron)) ft remove and convert to numeric
ft Sum data by municipio because some municipios are split among districts and/or are divided into multiple precincts
Electores <- tapply(xl2$"Electores", list(xl2$"CNTYIDFP"),sum) ft sum registered voters by municipio (Electores
lncritos==Registered Voters)
Electores <- data.f rame(rownames(Electores),Electores)
colnames(Electores) <- cf'CNTYIDFP", "Electores")
Votaron <-tapply(xl2$"Votaron", list(xl2$"CNTYIDFP"),sum) ft sum voted by municipio (Votaron==Voted)
Votaron <- data. frame(rownames(Votaron),Votaron)
colnames(Votaron) <- c("CNTYIDFP", "Votaron")
xl2 <- left_join(Electores, Votaron, by="CNTYIDFP")
xl2$Year <-"2012"
x <- xl2
ft Population 18+ by Municipio (5-yr estimates)
ft 2012 population based on 5-year estimates 2009 - 2013
xl2 <- read.csv("L:/Priv/Sustainable_Community_Projects/SanJuan/Projects/HWBI/Data_Downloads/Populationl
8+/PR_Municipio/ACS_13_5YR_DP05_with_ann.csv", header=T, stringsAsFactors = F)
xl2 <- xl2[,c("GEO.id2","HC01_VC32")] ft HC01_VC32 Estimate=SEX AND AGE -18 years and over
xl2$GEO.id2 <- as.character(xl2$GEO.id2)
colnames(xl2)[colnames(xl2)=="HC01_VC32"] <- "Populationl8"
xl2$Year <-"2012"
pop <- xl2
x <- left_join(x, pop, by=c("CNTYIDFP"="GEO.id2", "Year"))
x$REGVOTRS <- round((x$Electores/x$Populationl8)*100,2)
x$REGVOTRS <- ifelse((x$REGVOTRS>100),100,x$REGVOTRS) ft replace >100% voter registration with 100%
(Population estimate errors)
colnames(x)[colnames(x)=="REGVOTRS"] <- "REGVOTRS_m"
HWBI <- left_join(HWBI, x[,c("CNTYIDFP", "Year", "REGVOTRS_m")], by=c("CNTYIDFP", "Year"))
22
-------
Table 3.6. Sample of formatted county-level metric data for 2012 voter registration by
municipio. The HWBI metric REGVOTRS is calculated as the percentage of the population over
age 18 (Populationl8) registered to vote (Electores).
CNTYIDFP
Electores
Votaron
Year
Populationl8
REGVOTRS m
72001
14198
12134
2012
14520
97.78
72003
28613
24071
2012
32103
89.13
72005
35498
27702
2012
46127
76.96
72007
19381
15572
2012
21458
90.32
72009
19603
16143
2012
19583
100.00
72011
20085
16821
2012
22369
89.79
Example 3: Organizing and formatting metric data based on occurrence data
Some metrics were based only on available occurrence data that represent individual events
(e.g., mortalities) and needed to be summarized by spatial area. For example, diabetes
mortality (DIABMORT) was calculated as the percentage of total deaths using occurrence data
downloaded from the Centers for Disease Control and Prevention-National Center for Health
Statistics' National Vital Statistics System Multiple Cause of Death database (Centers for Disease
Control and Prevention, 2015e). These data were available for 2000-2004 by county and MSA
and 2000-2007 and 2010-2012 by state. In this example, we present the unprocessed 2000
mortality data (Table 3.7) where each row represents a mortality event by underlying cause of
death identified by state, city, and/or region. These datasets were summarized as the
percentage of total deaths attributed to diabetes by spatial scale (Table 3.8).
Box 3.6. R Code to read in diabetes mortality occurrence data for 2000.
# Read in diabetes mortality occurrence data for 2000
xOO <- read. csv("L:/Priv/Sustainable_Community_Projects/SanJuan/Projects/HWBI/Data_Downloads/CDC_Mortali
ty/mortterr2000.csv", header=T)
# pmsares (NCHS Primary/Metropolitan Statistical Area (P/MSA) of Residence) # 1990 PMSA descriptions=http://ww
w.cdc.gov/nchs/data/dvs/mcd/geog94msaposs.txt
# fipsctyr (County of Residence (FIPS))
# fipspmsa (PMSA/MSA of Residence (FIPS))
# fipscmsa (CMSA of Residence (FIPS)) # larger scale than MSA and PMSA (pmsares)
# ucrll3 (113 Cause Recode (A recode of the ICD cause code into 113 groups for NCHS publications))
# ucod (UNDERLYING CAUSE OF DEATH) frin ICD Code (10th Revision), See the International Classification of Diseases
, 1992 Revision, Volume 1
xOO <- x00[,c("year", "fipsstr", "fipsctyr", "pmsares", "fipspmsa", "fipscmsa", "ucrll3", "ucod")]
xOO <- x00[x00$fipsstr ==72, ] # Subset Puerto Rico
23
-------
Table 3.7. Sample of unformatted and unprocessed diabetes mortality occurrence data
downloaded from the Centers for Disease Control and Prevention-National Center for Health
Statistics' National Vital Statistics System Multiple Cause of Death database. Each row
represents an individual mortality by underlying cause of death (ucod) identified by state
(fipsstr), city (fipsctyr), and/or MSA (fipspmsa).
year
fipsstr
fipsctyr
pmsares
fipspmsa
fipscmsa
ucrll3
ucod
2000
72
72999
004
4840
0
63
1209
2000
72
72999
001
60
0
69
1120
2000
72
72999
001
60
0
67
1500
2000
72
72999
001
60
0
59
1219
2000
72
72999
001
60
0
70
164
2000
72
72999
001
60
0
59
1219
Box 3.7. R Code to calculate percentage of deaths attributed to diabetes.
# Calculate number of total mortality events by spatial scale
msaTotal <- tapply(x00$"ucrll3", list(xOO$"pmsares"),length) # total mortality events by MSA
msaTotal <- data.f rame(rownames(msaTotal),msaTotal)
colnames(msaTotal) <- c("pmsares", "AllCauses")
mTotal <- tapply(x00$"ucrll3", list(xOO$"fipsctyr"),length) # total mortality events by municipio
mTotal <- data.f rame(rownames(mTotal),mTotal)
colnames(mTotal) <- cf'CNTYIDFP", "AllCauses")
# Calculate MSA-level diabetes mortality as a percentage of total mortality
yOO <- x00[x00$ucrll3=="46",] #Subset diabetes mortality using 113 Cause Recode
msa <- tapply(y00$"ucrll3", list(yOO$"pmsares"),length) # diabetes mortality by MSA
msa <- data.f rame(rownames(msa),msa)
colnames(msa) <- c("pmsares", "y")
msa$Year <- "2000"
msa <- left_join(msa, msaTotal, by=" pmsares")
msa$DIABMORT_msa <- ifelse((msa$pmsares!="000" & msa$y >=50), round((msa$y/msa$AIICauses)*100,2),
NA)
z <- msa[(msa$pmsares=="000" | msa$y<50), ] #PR-(MSA with n>=50)
msa$DIABMORT_PR <-ifelse((msa$pmsares=="000" | msa$y<50), round(((sum(z$y,
na. rm=T))/(sum(z$AIICauses, na. rm=T)))*100,2), NA) # "PR-msa"
DIABMORT_msa <- msa[,c("pmsares", "Year", "DIABMORT_msa", "DIABMORT_PR")]
continued next page
24
-------
Box 3.7. continued
ft Calculate municipio-level diabetes mortality as a percentage of total mortality
m <- tapply(y00$"ucrll3", list(yOO$"fipsctyr"),length) ft diabetes mortality by municipio
m <- data.frame(rownames(m),m)
colnames(m) <- cf'CNTYIDFP", "y")
m$Year <- "2000"
m <- left_join(m, mTotal, by="CNTYIDFP")
m$DIABMORT_m <- ifelse((m$CNTYIDFP!="72999" & m$y >=50), round((m$y/m$AIICauses)*100,2), NA)
z <- m[(m$CNTYIDFP=="72999" | m$y<50), ] ft PR-m with n>=50
m$DIABMORT_PR <-ifelse((m$CNTYIDFP=="72999" | m$y<50), round(((sum(z$y, na. rm=T))/(sum(z$AIICauses,
na. rm=T)))*100,2), NA) ft "PR-m"
DIABMORT_m<- m[,cf'CNTYIDFP", "Year", "DIABMORT_m", "DIABMORT_PR")]
ft Join diabetes mortality data to county x year template
colnames(DIABMORT_m) <- c("CNTYIDFP","Year","DIABMORT_m","DIABMORT_PR.m")
colnames(DIABMORT_msa) <- c("pmsares","Year","DIABMORT_msa","DIABMORT_PR.msa")
ft Convert MSA codes for 2003-2004 from FIPS to Vital Statistics Codes
DIABMORT_msa$pmsares <- with(DIABMORT_msa, ifelse((Year=="2003" | Year=="2004") & pmsares=="0", "000
ifelse((Year=="2003" | Year=="2004") & pmsares=="60", "001",
ifelse((Year=="2003" | Year=="2004") & pmsares=="470", "002",
ifelse((Year=="2003" | Year=="2004") & pmsares=="1310", "003",
ifelse((Year=="2003" | Year=="2004") & pmsares=="4840", "004",
ifelse((Year=="2003" | Year=="2004") & pmsares=="6360", "005",
ifelse(((Year=="2003" | Year=="2004") & pmsares=="7440"), "006", pmsares))))))))
mort <- HWBI[,c("CNTYIDFP","Year")]
ft Add pmsa to lookup table & HWBI
lookupM$pmsa_1990 <- ifelse(lookupM$MSA_00=="Aguadilla", "001",
ifelse(lookupM$MSA_00=="Arecibo", "002",
ifelse(lookupM$MSA_00=="Caguas", "003",
ifelse(lookupM$MSA_00=="Mayaguez", "004",
ifelse(lookupM$MSA_00=="Ponce", "005",
ifelse((lookupM$MSA_00=="San Juan"), "006", NA))))))
x <- lookupM[,c("CNTYIDFP", "pmsa_1990")]
mort <- left_join(mort, x, by="CNTYIDFP")
ft Municipio
x <- DIABMORT_m[,c("CNTYIDFP","Year","DIABMORT_m")]
mort <- left_join(mort, x, by=c("CNTYIDFP", "Year")) ft Join to master dataframe
ft PR-municipio
x <-tapply(DIABMORT_m$"DIABMORT_PR.m", list(DIABMORT_m$"Year"),mean, na.rm=T)
x <- data. f rame(rownames(x),x)
colnames(x) <- c("Year", "DIABMORT_PR.m")
mort <- left_join(mort, x, by="Year") ft Join to master dataframe
continued next page
25
-------
Box 3.7. continued
ft MSA
x <- DIABMORT_msa[,c("pmsares","Year","DIABMORT_msa")]
x$pmsares <- as.character(x$pmsares)
mort <- left_join(mort, x, by=c("pmsa_1990"="pmsares", "Year")) ft Join to master dataframe
ft PR-msa
x <- tapply(DIABMORT_msa$"DIABMORT_PR.msa", list(DIABMORT_msa$"Year"),mean, na. rm=T)
x <- data. f rame(rownames(x),x)
colnames(x) <- c("Year", "DIABMORT_PR.msa")
mort <- left_join(mort, x, by="Year") ft Join to master dataframe
Table 3.8. Sample of formatted county-level metric data for 2000 diabetes mortality
(percentage) by municipio for available spatial scales, including municipio (DIABMORT_m),
state without county-level data (DIABMORT_PR.m), MSA (DIABMORT_msa), and state
without MSA-level data (DIABMORT_PR.msa).
CNTYIDFP
Year
pmsa_1990
DIABMORTm
DIABMORTPR.m
DIABMORTmsa
DIABMORTPR.msa
72001
2000
NA
NA
8.58
NA
9.2
72003
2000
001
NA
8.58
NA
9.2
72005
2000
001
NA
8.58
NA
9.2
72007
2000
006
NA
8.58
NA
9.2
72009
2000
NA
NA
8.58
NA
9.2
72013
2000
002
NA
8.58
NA
9.2
3.4. Merging Metric Data by municipio x year Combination
We merged data from multiple sources with unique formatting into a single table with all
metrics and available geographies and organized by municipio x year combination (Table 3.9).
Although this organization is set up for processing and calculation in R, additional database or
data processing software could be employed. Metrics were named with geographic suffixes to
identify spatial scale of original data source (e.g., Commonwealth of Puerto Rico indicated by
METRICNAME_pr, municipio indicated by METRICNAMEjm, PUMA delineated by 2000 U.S.
Census indicated by METRICNAME_p00).The dataset was organized in a wide format, where
each row represents a municipio x year combination and columns contain metric data by
county for all available spatial scales. In the next section, metric data for only the finest
available spatial scale were selected for each metric x county x year according to spatial
hierarchy.
26
-------
Table 3.9. Sample of formatted county-level metric data for San Juan municipio merged by
municipio and year for available spatial scales. Metrics were named with geographic suffixes
to identify spatial scale of original data source.
HOMEVAL
HOMEVAL
HOMEVAL REGVOTRS
DIABMORT
DIABMORT
DIABMORT
DIABMORT
DIABMORT
NAME
CNTYIDFP
Year
_m
_p00
_pl0
_m
_m
PR.m
_msa
PR.msa
_t
San Juan
72127
2000
NA
NA
NA
NA
7.25
8.58
7.47
9.20
NA
San Juan
72127
2001
NA
NA
NA
NA
7.46
8.91
7.60
9.32
NA
San Juan
72127
2002
NA
NA
NA
NA
7.40
9.27
7.86
9.17
NA
San Juan
72127
2003
NA
NA
NA
NA
7.40
9.38
8.07
10.00
NA
San Juan
72127
2004
NA
NA
NA
78.07
8.71
9.75
8.24
10.55
NA
San Juan
72127
2005
148100
146800
NA
NA
NA
NA
NA
NA
9.27
San Juan
72127
2006
150300
150200
NA
NA
NA
NA
NA
NA
9.03
San Juan
72127
2007
164200
167500
NA
NA
NA
NA
NA
NA
9.17
San Juan
72127
2008
173400
171700
NA
77.76
NA
NA
NA
NA
NA
San Juan
72127
2009
170500
167900
NA
NA
NA
NA
NA
NA
NA
San Juan
72127
2010
168900
163500
NA
NA
NA
NA
NA
NA
9.84
San Juan
72127
2011
164800
159800
NA
NA
NA
NA
NA
NA
10.54
San Juan
72127
2012
163800
NA
158500
76.70
NA
NA
NA
NA
10.14
San Juan
72127
2013
164400
NA
159100
NA
NA
NA
NA
NA
NA
3.5. Selecting the Finest Available Spatial Scale for Each metric x county
xyear Combination
Although Puerto Rico HWBI was developed at the municipio-level, not all metric data were
available at that spatial scale. In these cases, data was organized hierarchically according to
spatial area (Figure 3.2). If available, we first selected county data (municipio), then data
aggregated by statistical area (PUMA or MSA), then other regional data (World Values Survey
Regions), then state data (Commonwealth of Puerto Rico), and lastly U.S. territory data
(including Puerto Rico, U.S. Virgin Islands, Guam, American Samoa, and Northern Mariana
Islands, whereby Puerto Rico comprised approximately 95% of observations).
27
-------
Figure 3.2. Summary of spatial hierarchy used to select data for Puerto Rico HWBI.
Are data available by
municlpio?
¦No-
-Yes-
Are data available by
U.S. Census Bureau
statistical area?
County
-Yes-
Are data available by U.S.
Office of Management and
Budget statistical area?
-Yes-
to 1
y
-Yes-
World Values Survey
(WVS) region
Yes
-NOr
-No [
-Yes-
Yes
No
State
Public Use
Microdata Area
(PUMA)
Are data available by
World Values Survey
region?
Are date available for
Commonwealth of
Puerto Rico?
lO—1—"t
4?
State-municipios or
State-region
No data avails bk
Is it possible to subtract
large municipios or
regions with available
data from
Commonwealth?
Are data available tor U.S.
territories (Puerto Rico, U.S.
Virgin islands, Guam,
AmerfcanSamoa, Northern
Mariana i stands)?
28
-------
The dataset was converted from wide (Tables 3.10 and 3.11) to long (Table 3.12) format to
facilitate selection of the finest available spatial scale for each metric x county x year
combination. This allowed for the creation of a single variable (datascale) to describe the spatial
scale of each metric. Here we present results for rural Adjuntas and urban San Juan municipios
using median home value and diabetes mortality metric data.
Table 3.10. Sample of wide-format county-level median home value (HOMEVAL) metric data
for all available spatial scales. Data are included for densely populated San Juan municipio and
less populated Adjuntas municipio.
NAME
CNTYIDFP
Year
HOMEVALm
HOMEVAL pOO HOMEVAL_plO
Adjuntas
72001
2000
NA
NA
NA
Adjuntas
72001
2001
NA
NA
NA
Adjuntas
72001
2002
NA
NA
NA
Adjuntas
72001
2003
NA
NA
NA
Adjuntas
72001
2004
NA
NA
NA
Adjuntas
72001
2005
NA
83200
NA
Adjuntas
72001
2006
NA
76200
NA
Adjuntas
72001
2007
NA
86000
NA
Adjuntas
72001
2008
NA
98400
NA
Adjuntas
72001
2009
NA
98300
NA
Adjuntas
72001
2010
NA
97100
NA
Adjuntas
72001
2011
NA
116700
NA
Adjuntas
72001
2012
NA
NA
89500
Adjuntas
72001
2013
NA
NA
86400
San Juan
72127
2000
NA
NA
NA
San Juan
72127
2001
NA
NA
NA
San Juan
72127
2002
NA
NA
NA
San Juan
72127
2003
NA
NA
NA
San Juan
72127
2004
NA
NA
NA
San Juan
72127
2005
148100
146800
NA
San Juan
72127
2006
150300
150200
NA
San Juan
72127
2007
164200
167500
NA
San Juan
72127
2008
173400
171700
NA
San Juan
72127
2009
170500
167900
NA
San Juan
72127
2010
168900
163500
NA
San Juan
72127
2011
164800
159800
NA
San Juan
72127
2012
163800
NA
158500
San Juan
72127
2013
164400
NA
159100
29
-------
Table 3.11. Sample of wide-format county-level diabetes mortality (DIABMORT) metric data
for all available spatial scales. Data are included for densely populated San Juan municipio and
less populated Adjuntas municipio.
NAME
CNTYIDFP
Year
DIABMORT m
DIABMORT PR.m
DIABMORT msa
DIABMORT PR.msa
DIABMORT t
Adjuntas
72001
2000
NA
8.58
NA
9.20
NA
Adjuntas
72001
2001
NA
8.91
NA
9.32
NA
Adjuntas
72001
2002
NA
9.27
NA
9.17
NA
Adjuntas
72001
2003
NA
9.38
NA
10.00
NA
Adjuntas
72001
2004
NA
9.75
NA
10.55
NA
Adjuntas
72001
2005
NA
NA
NA
NA
9.27
Adjuntas
72001
2006
NA
NA
NA
NA
9.03
Adjuntas
72001
2007
NA
NA
NA
NA
9.17
Adjuntas
72001
2008
NA
NA
NA
NA
NA
Adjuntas
72001
2009
NA
NA
NA
NA
NA
Adjuntas
72001
2010
NA
NA
NA
NA
9.84
Adjuntas
72001
2011
NA
NA
NA
NA
10.54
Adjuntas
72001
2012
NA
NA
NA
NA
10.14
Adjuntas
72001
2013
NA
NA
NA
NA
NA
San Juan
72127
2000
7.25
8.58
7.47
9.20
NA
San Juan
72127
2001
7.46
8.91
7.60
9.32
NA
San Juan
72127
2002
7.40
9.27
7.86
9.17
NA
San Juan
72127
2003
7.40
9.38
8.07
10.00
NA
San Juan
72127
2004
8.71
9.75
8.24
10.55
NA
San Juan
72127
2005
NA
NA
NA
NA
9.27
San Juan
72127
2006
NA
NA
NA
NA
9.03
San Juan
72127
2007
NA
NA
NA
NA
9.17
San Juan
72127
2008
NA
NA
NA
NA
NA
San Juan
72127
2009
NA
NA
NA
NA
NA
San Juan
72127
2010
NA
NA
NA
NA
9.84
San Juan
72127
2011
NA
NA
NA
NA
10.54
San Juan
72127
2012
NA
NA
NA
NA
10.14
San Juan
72127
2013
NA
NA
NA
NA
NA
Box 3.8. R Code to reformat data from wide to long.
preData <- read.csv("./Data_Downloads/BackupHWBIall.csv", header=T, stringsAsFactors = F) ffthis is the
organized
HWBI data in wide format with geographic identifier concatenated with metric identifier
x <- gather(preData, "Metric", "Value", 4:ncol(preData), na. rm=T) # create long format of organized data
x <- x %>% separatef'Metric", into = c("METRIC_VAR","geography"), sep = #separate metric and geograp
hy into separate variables
# Identify data scale and collapse synonymous spatial scales (e.g., pOO and plO are both PUMA-level geography)
x$datascale <-with(x, ifelse(geography=="wvsR", "WVS Region",
ifelse(geography=="msa", "MSA",
ifelse(geography=="pOO" | geography=="plO" | geography=="p", "PUMA",
ifelse((geography=="m" | geography=="m.all"), "County", geography)))))
30
-------
Table 3.12. Sample of long-format county-level metric data including data for all available
spatial scales.
NAME CNTYIDFP Year METRICVAR geography Value datascale
San Juan 72127 2004 DIABMORT
San Juan 72127 2004 DIABMORT
San Juan 72127 2004 DIABMORT
San Juan 72127 2004 DIABMORT
m 8.71 County
PR.m 9.75 PR-municipio
msa 8.24 MSA
PR.msa 10.55 PR-MSA
Example 4: Selecting finest spatial scale for metric data based on multiple spatial scales
A sample of data are presented for HOMEVAL for the densely populated San Juan municipio
and more rural Adjuntas municipio to illustrate working with different spatial scales. Home
value data were collected for three spatial scales: municipio (HOMEVAL_m), 2000 Census
boundary PUMAs (HOMEVAL_pOO), and 2010 Census boundary PUMAs (HOMEVAL_plO). In
order to calculate HWBI at the county-level, metric data needed to be reduced to a single
measure for each metric x county x year combination. Although municipio-specific data were
prioritized, sometimes this was not the finest available spatial scale. In these cases, metric data
were selected to include only the finest available spatial scale for each metric x county x year
combination. In this example, municipio-level data were selected if available, otherwise PUMA-
level data were used according to year. For some combinations, no data was available and were
imputed in a later data processing step (Chapter 4).
Example 5: Selecting finest spatial scale for metric data based on more complex spatial scales
Diabetes mortality (DIABMORT) data was collected for a more complex set of spatial scales
because it was based on raw occurrence data. In addition to collecting data for multiple spatial
scales, we were able to calculate mortality for the Commonwealth of Puerto Rico without
including cases for large counties and MSAs (i.e. Commonwealth mortality outside of large
municipios and metropolitan areas). This was possible for most occurrence data (e.g., number
of deaths) where we could subtract large municipios or MSAs with available data from the
collective data to get a metric value for all remaining municipios or MSAs. Data was collected by
municipio (DIABMORT_m), Commonwealth of Puerto Rico without municipio data
(DIABMORT_PR.m), MSA (DIABMORT_msa), Commonwealth of Puerto Rico calculated without
MSA data (DIABMORT_PR.msa), and all U.S. territories (DIABMORT_t). The need to collect data
for multiple spatial scales and then select the finest available is evident in Table 3.11 where
data are again presented for rural Adjuntas and urban San Juan municipios.
31
-------
During the selection process, data was subset by spatial scale and recombined in order to apply
the decision hierarchy to select only the finest available spatial scale for each metric x county x
year combination (Table 3.13).
Box 3.9. R Code to select data at the finest available spatial scale.
xCounty <- x[x$datascale=="County", ]
colnames(xCounty)[colnames(xCounty)=="Value"] <- "valueCounty"
colnames(xCounty)[colnames(xCounty)=="datascale"] <- "datascaleCounty"
xPUMA <- x[x$datascale=="PUMA", ]
colnames(xPUMA)[colnames(xPUMA)=="Value"] <- "valuePUMA"
colnames(xPUMA)[colnames(xPUMA)=="datascale"] <- "datascalePUMA"
xMSA <- x[x$datascale=="MSA", ]
colnames(xMSA)[colnames(xMSA)=="Value"] <- "valueMSA"
colnames(xMSA)[colnames(xMSA)=="datascale"] <- "datascaleMSA"
xWVS <- x[x$datascale=="WVS Region", ]
colnames(xWVS)[colnames(xWVS)=="Value"] <- "valueWVS"
colnames(xWVS)[colnames(xWVS)=="datascale"] <- "datascaleWVS"
xPR <- x[x$datascale=="PR", ]
colnames(xPR)[colnames(xPR)=="Value"] <- "valuePR"
colnames(xPR)[colnames(xPR)=="datascale"] <- "datascalePR"
xTerritory <- x[x$datascale=="t", ]
colnames(xTerritory)[colnames(xTerritory)=="Value"] <- "valueTerritory"
colnames(xTerritory)[colnames(xTerritory)=="datascale"]<-"datascaleTerritory"
xPR.msa <- x[x$datascale=="PR.msa", ]
colnames(xPR.msa)[colnames(xPR.msa)=="Value"] <- "valuePR.msa"
colnames(xPR.msa)[colnames(xPR.msa)=="datascale"] <- "datascalePR.msa"
xPR.m <- x[x$datascale=="PR.m", ]
colnames(xPR.m)[colnames(xPR.m)=="Value"] <- "valuePR.m"
colnames(xPR.m)[colnames(xPR.m)=="datascale"] <- "datascalePR.m"
continued next page
32
-------
Box 3.9. continued.
# Create template of all metric x county x year combinations
skeleton <- preData[preData$Year!=2014 ,c("CNTYIDFP", "Year")] #remove 2014
lookupMetrics <- read.csv("./PR_setup_lookup.csv", header=T, stringsAsFactors = F) #look up table of HWBI
metric
descriptions
metrics <- lookupMetrics[!is. na(lookupMetrics$PR_modification), "METRIC_VAR"] # subset of metrics available for
Puerto Rico
skeleton <- merge(skeleton, metrics)
colnames(skeleton)[colnames(skeleton)=="y"] <- "METRIC_VAR"
skeleton$METRIC_VAR <- as.character(skeleton$METRIC_VAR)
y <- left_join(skeleton, xCounty, by=c("CNTYIDFP", "Year", "METRIC_VAR"))
y <- left_join(y, xPUMA, by=c("CNTYIDFP", "Year", "METRIC_VAR"))
y <- left_join(y, xMSA, by=c("CNTYIDFP", "Year", "METRIC_VAR"))
y <- left_join(y, xWVS, by=c("CNTYIDFP", "Year", "METRIC_VAR"))
y <- left_join(y, xPR, by=c("CNTYIDFP", "Year", "METRIC_VAR"))
y <- left_join(y, xTerritory, by=c("CNTYIDFP", "Year", "METRIC_VAR"))
y <- left_join(y, xPR.m, by=c("CNTYIDFP", "Year", "METRIC_VAR"))
y <- left_join(y, xPR.msa, by=c("CNTYIDFP", "Year", "METRIC_VAR"))
# ORIG_MEASURE = value of finest available spatial scale
y$ORIG_MEASURE <- with(y, ifelse(!is. na(valueCounty), valueCounty,
....ifelse(!is.na(valuePUMA), valuePUMA,
ifelse(!is.na(valueMSA), valueMSA,
ifelse(!is.na(valueWVS), valueWVS,
ifelse(!is.na(valuePR.msa), valuePR.msa,
ifelse(!is.na(valuePR.m), valuePR.m,
ifelse(!is.na(valuePR), valuePR,
ifelse((!is. na(valueTerritory)), valueTerritory, NA)))))))))
# datascale = datascale of finest available spatial scale
y$datascale <- with(y, ifelse(!is. na(datascaleCounty), datascaleCounty,
ifelse(!is.na(datascalePUMA), datascalePUMA,
ifelse(!is. na(datascaleMSA), datascaleMSA,
ifelse(!is.na(datascaleWVS), datascaleWVS,
ifelse(!is. na(datascalePR.msa), datascalePR.msa,
ifelse(!is.na(datascalePR.m), datascalePR.m,
ifelse(!is. na(datascalePR), datascalePR,
ifelse((!is. na(datascaleTerritory)), datascaleTerritory, NA)))))))))
# Collapse synonymous spatial scales in spatial scale hierarchy
y$datascale <-with(y, ifelse((datascale=="PR" | datascale=="t" | datascale=="PR.msa" | datascale=="PR.m"),
"State", datascale))
MasterHWBIallGeo <-y[ ,c("CNTYIDFP","Year","METRIC_VAR","ORIG_MEASURE","datascale")] # Co//opse
synonymous spatial scales in spatial scale hierarchy
33
-------
Table 3.13. Sample of output data with finest available spatial scale selected for each metric x
county x year combination for the densely populated San Juan municipio and less populated
Adjuntas municipio.
CNTYIDFP
NAME
Year
METRIC VAR
ORIG MEASURE
datascale
72001
Adiuntas
2000
DIABMORT
9.20
State
72001
2001
DIABMORT
9.32
State
72001
2002
DIABMORT
9.17
State
72001
2003
DIABMORT
10.00
State
72001
2004
DIABMORT
10.55
State
72001
2005
DIABMORT
9.27
State
72001
2006
DIABMORT
9.03
State
72001
2007
DIABMORT
9.17
State
72001
2008
DIABMORT
NA
NA
72001
2009
DIABMORT
NA
NA
72001
2010
DIABMORT
9.84
State
72001
2011
DIABMORT
10.54
State
72001
2012
DIABMORT
10.14
State
72001
2013
DIABMORT
NA
NA
72127
San Juan
2000
DIABMORT
7.25
Countv
72127
2001
DIABMORT
7.46
Countv
72127
2002
DIABMORT
7.40
Countv
72127
2003
DIABMORT
7.40
Countv
72127
2004
DIABMORT
8.71
Countv
72127
2005
DIABMORT
9.27
State
72127
2006
DIABMORT
9.03
State
72127
2007
DIABMORT
9.17
State
72127
2008
DIABMORT
NA
NA
72127
2009
DIABMORT
NA
NA
72127
2010
DIABMORT
9.84
State
72127
2011
DIABMORT
10.54
State
72127
2012
DIABMORT
10.14
State
72127
2013
DIABMORT
NA
NA
72001
2000
HOMEVAL
NA
NA
72001
2001
HOMEVAL
NA
NA
72001
2002
HOMEVAL
NA
NA
72001
2003
HOMEVAL
NA
NA
72001
2004
HOMEVAL
NA
NA
72001
2005
HOMEVAL
83200.00
PUMA
72001
2006
HOMEVAL
76200.00
PUMA
72001
2007
HOMEVAL
86000.00
PUMA
72001
2008
HOMEVAL
98400.00
PUMA
72001
2009
HOMEVAL
98300.00
PUMA
72001
2010
HOMEVAL
97100.00
PUMA
72001
2011
HOMEVAL
116700.00
PUMA
72001
2012
HOMEVAL
89500.00
PUMA
72001
2013
HOMEVAL
86400.00
PUMA
72127
2000
HOMEVAL
NA
NA
72127
2001
HOMEVAL
NA
NA
72127
2002
HOMEVAL
NA
NA
72127
2003
HOMEVAL
NA
NA
72127
2004
HOMEVAL
NA
NA
72127
2005
HOMEVAL
148100.00
Countv
72127
2006
HOMEVAL
150300.00
Countv
72127
2007
HOMEVAL
164200.00
Countv
72127
2008
HOMEVAL
173400.00
Countv
72127
2009
HOMEVAL
170500.00
Countv
72127
2010
HOMEVAL
168900.00
Countv
72127
2011
HOMEVAL
164800.00
Countv
34
-------
Chapter 4: Processing Data for HWBI Calculation
4.1. Summary of Data Processing Prior to HWBI Calculation
Raw metric data require processing before HWBI calculation. This includes eliminating all but
the finest available spatial scale, imputing missing data to create a temporally and spatially
complete dataset, and standardizing data for comparison among metrics and populations. Here
we describe two distinct phases of data preparation: formatting and processing. Formatting
involves structuring rows and columns of the dataset which can be accomplished through
various programming or database functions. Processing refers to a four-step workflow of HWBI-
specific methods to impute and standardize data.
Although processing steps are the same for overall HWBI calculation, specific methods differ
according to user needs and application of the index. Therefore users should consider how the
index will be used as a prerequisite for data processing and HWBI calculation. For example,
determining whether an HWBI can be compared among populations depends on how data are
normalized and standardized. Imputation methods may also vary based on the need to fill in
missing data. Table 4.1 summarizes five unique versions of the HWBI developed during the
adaptation for Puerto Rico. These vary in terms of the number of included metrics, how metrics
were normalized and standardized, how missing metrics were imputed, and the source of
imputation data. Versions are compared in Chapters 6 and 8 to illustrate differences that result
from method selection. In the sample R code shown here and in Chapter 5, we calculated HWBI
using "U.S. imputations for only missing indicators" to compare among Puerto Rico municipios
and to U.S. counties.
Table 4.1. Summary of HWBI versions for Puerto Rico using alternative processing methods.
Version
Number of
metrics
Normalization and
standardization source
Imputation
methods
Imputation
source
Within Puerto Rico
74
Puerto Rico, unless single
observation then U.S.
None
NA
U.S. imputations for all
missing metrics
80
U.S.
All missing metrics
U.S.
Hawai'i imputations for all
missing metrics
80
U.S.
All missing metrics
Hawai'i
U.S. imputations for only
missing indicators
72
U.S.
Only if metric removal
resulted in indicator loss
U.S.
Hawai'i imputations for only
missing indicators
72
U.S.
Only if metric removal
resulted in indicator loss
Hawai'i
35
-------
I f i','cessing Metric Data Pri m i H FI i I. ulation
Spatial continuity of metric data, especially considered over time, is often incomplete. Of the 80
metrics used to calculate HWBI, few were available annually at the municipio level for all 78
municipios. Therefore imputation methods were employed to fill in missing years of data
(temporal imputation) and fill in missing county-level data (spatial imputation) using
summarized data from coarser spatial scales. Spatial imputation was accomplished primarily
using data from within Puerto Rico when available, otherwise U.S. data were used. Additionally,
metrics are based on a variety of units of measure that cannot be directly compared without
standardization. Thus processing steps were performed to create a temporally and spatially
complete data set with metrics normalized and standardized for HWBI calculation (Figure 4.1).
Temporal imputation
Geographic imputation
Metric data for
smallest available geography
for available years
Metric data for
smallest available geography
for all years
Metric data for
all municipios
for all years
Normalization
Normalized metric data for
all municipios
for all years
\ | J
Standardization
Standardized metric data for
all municipios
for all years
\ )
Figure 4.1. Summary of processing steps to prepare data for HWBI calculation.
36
-------
Here we describe processing as an HWBI-specific workflow with four steps: temporal
imputation, creation of fill data, spatial imputation, and standardization. We demonstrated
these processing steps in R using the previously described packages in addition to Hoo' version
1.7.12 (Zeileis et al., 2015) for temporal imputation.
Processing step 1: Correct data structure and temporal imputation
This processing step constructs a county-level metric template, identifies original metric data source and
populates template.
Step lo. Construct o county-level metric dotoset (metric x county x year)
Input data include downloaded metric data organized by county (Table 3.13), lookup table for
Puerto Rico spatial scales within spatial hierarchy (Table 4.2), and lookup table of HWBI metric
descriptions (Table 4.3). We constructed a template listing the metrics available for Puerto Rico
HWBI for each municipio by year (Table 4.4). This dataset formed the template for HWBI
processing to which additional variables were joined or calculated.
Table 4.2. Sample of lookup table for Puerto Rico spatial scales within spatial hierarchy.
FIPS Year
SCALE PUMAREGION MSA
REGION WVS_
REGION
STATE ABBR COUNTY POPULATION
72 2000
State NA NA
NA
PR NA
3808605
1 2000
WVS Region NA NA
Centra
PR NA
1096642
100 2000
MSA Region NA Aguadilla NA
PR NA
146505
72003 2000
County 7200100 Aguadilla Oeste
PR Aguada 42074
Table 4.3. Sample of lookup table of HWBI metric descriptions.
POS_NEG_
DOMAIN
INDICATOR
METRICVAR
METRIC
units
PRmodification
Connection to Biophilia
BEAUSPRT
P
Proportion
NA
Nature
Connection to Biophilia
ALLOFLFE
P
Percentage
Substitute measure
Nature
Cultural
Cultural Activity Participation
TOTRATE
P
Rate
Modified source
Fulfillment
Cultural
Cultural Activity Participation
PERARTS
P
Rate
Substitute measure
Fulfillment
and units
Education
Basic Educational Knowledge
MATHTEST
P
Percentage
Modified source
and Skills of Youth
Education
Basic Educational Knowledge
READTEST
P
Percentage
Modified source
and Skills of Youth
37
-------
Box 4.1. R Code to load input data and get unique metric variables with data available for
Puerto Rico for each municipio and year.
ft Load packages and input datasets for processing
library(tidyr)
library(dplyr)
library(zoo)
dataPR <- read.csv("./PR_MasterHWBIallCases.csv", header=T, stringsAsFactors=F) ft pre-processed data output
lookupFIPS <- read.csv("./PR_fipsdata.csv", header=T, stringsAsFactors=F) ft lookup table for Puerto Rico spatial
scales within spatial hierarchy
lookupMetrics <- read.csv("./PR_setup_lookup.csv", header=T, stringsAsFactors=F) ft lookup table for HWBI metric
descriptions
meanOutUS<- read.csv("./US_MEANOUT.csv", header=T, stringsAsFactors=F) ft interguartile ranges for U.S. metrics
for normalization
minmaxFinalUS <- read.csv("./US_MINMAXFINAL.csv", header=T, stringsAsFactors=F) ft summary of U.S. metrics for
standardization
USdomMetricsFillRG <- read.csv("./dom_metrics_fill_rg.csv", header=T, stringsAsFactors=F) ftfill data by RUCC-GINI
calculated for the U.S.
ft Get unigue metric variables with data available for Puerto Rico and repeat for all municipios by year
uniqueMetrics <- dataPR[with(dataPR, do.call(order, list(METRIC_VAR))), ] ft get unigue metrics
uniqueMetrics <- do.call(rbind, by(uniqueMetrics, list(uniqueMetrics$METRIC_VAR), FUN=function(x) head(x, 1)))
uniqueFips <- lookupFIPS[lookupFIPS$SCALE=="County", c("FI PS","Year")] ft get unigue municipios by year
metricsComplete <- merge(uniqueMetrics[ ,c("METRIC_VAR","DOMAIN","INDICATOR","POS_NEG_METRIC")],
uniqueFips) ft get complete metric x county x year template
Table 4.4. Sample of county-level metric dataset for Puerto Rico where each row is a unique
metric x county x year combination. See Appendix B for variable descriptions.
METRICVAR
DOMAIN
INDICATOR
FIPS
Year
ACCMM
Safety and Security
Actual Safety
72001
2000
ADLTASTHMA
Health
Physical and Mental Health Conditions
72001
2000
ADULTLIT
Education
Participation and Attainment
72001
2000
ALCOHOL
Health
Lifestyle and Behavior
72001
2000
ALLOFLFE
Connection to Nature
Biophilia
72001
2000
ASTHMORT
Health
Life Expectancy and Mortality
72001
2000
38
-------
Step lb. Identify original metric data source and populate template
For each metric x county x year combination, the original metric data source for the finest
available spatial scale was identified according to spatial hierarchy (Figure 3.2) and joined to
the county-level metric template (Table 4.5).
Box 4.2. R Code to subset data by finest available spatial scale and join to the county-level
metric template.
ft Remove missing metric data
inmetrics <- dataPR[lis.na(dataPR$ORIG_MEASURE), ] ft remove missing values
inmetrics$measure <- inmetrics$ORIG_MEASURE ft new column name (measure)= available data by FIPS by year for
finest spatial scale available
ft Convert percentage to proportion
inmetrics$measure <- with(inmetrics, ifelse((units=="Percentage"), measure/100, measure))
inmetrics$units <- with(inmetrics, ifelse((units=="Percentage"), "Proportion", units)) ft update "units"
ft Get the number of non-missing values by metric and county
metricfipsn <- inmetrics %>%
group_by(METRIC_VAR, FIPS) %>%
summarise(metricfipsn=length(measure)) ft sample size by metric and county
ft Subset data by finest available spatial scale in order to assign original metric data source (ORIG_FIPS)
inmetrics <- leftJoin(inmetrics, metricfipsn, by=c("METRIC_VAR", "FIPS")) ft merge data sets
metrics <- inmetrics[ ,c("METRIC_VAR","FIPS","Year","measure","metricfipsn", "datascale")] ftselect variables
cmetrics <- metrics[metrics$datascale=="County", ] ftsubset county-level
cmetrics$ORIG_FIPS <- cmetrics$FIPS ft add ORIG_FIPS according to datascale (county FIPS code)
pmetrics <- metrics[metrics$datascale=="PUMA", ] ftsubset PUMA-level
lookupCounty <- lookupFIPS[lookupFIPS$SCALE=="County", c("FIPS","Year","PUMA_REGION")] ft create lookup table
with PUMA Region by county
pmetrics <- leftJoin(pmetrics, lookupCounty, by=c("FIPS", "Year"))
pmetrics$ORIG_FIPS <- pmetrics$PUMA_REGION ft add ORIG_FIPS according to datascale (PUMA region FIPS code)
mmetrics <- metrics[metrics$datascale=="MSA", ] ft subset MSA-level
lookupCounty <- lookupFIPS[lookupFIPS$SCALE=="County", cf'FIPS", "Year", "MSA_REGION")] ft create lookup table
with MSA Region by county
lookupMSA <- lookupFIPS[lookupFIPS$SCALE=="MSA Region", cf'FIPS", "Year","MSA_REGION")] ft create lookup
table for MSA geography
lookupMSA$MSA_FIPS <- lookupMSA$FIPS
lookupMSA$FIPS <- NULL
lookupCounty <- leftJoin(lookupCounty,lookupMSA, by=c("MSA_REGION","Year"))
lookupCounty$MSA_REGION <- NULL
mmetrics <- leftJoin(mmetrics, lookupCounty, by=c("FIPS", "Year"))
mmetrics$ORIG_FIPS <- mmetrics$MSA_FIPS ft add ORIG_FIPS according to datascale (MSA region FIPS code)
Continued on next page
39
-------
Box 4.2. Continued.
rmetrics <- metrics[metrics$datascale=="WVS Region", ] ffsubset WVS-level
lookupCounty <- lookupFIPS[lookupFIPS$SCALE=="County", cf'FIPS", "Year", "WVS_REGION")] ffcreate lookup table
with WVS Region by county
lookupWVS <- lookupFIPS[lookupFIPS$SCALE=="WVS Region", cf'FIPS", "Year", "WVS_REGION")] ff create lookup
table for WVS geography
lookupWVS$WVS_FIPS <- lookupWVS$FIPS
lookupWVS$FIPS <- NULL
lookupCounty <- leftJoin(lookupCounty,lookupWVS,by=c("WVS_REGION","Year"))
lookupCounty$WVS_REGION <- NULL
rmetrics <- leftJoin(rmetrics, lookupCounty, by=c("FlPS", "Year"))
rmetrics$ORIG_FIPS <- rmetrics$WVS_FIPS ft add ORIG_FIPS according to datascale (WVS region FIPS code)
smetrics <- metrics[metrics$datascale=="State", ] ff subset state-level
lookupCounty <- lookupFIPS[lookupFIPS$SCALE=="State", cf'FIPS", "Year")] ff create lookup table with state by
county
lookupCounty$ORIG_FIPS <- lookupCounty$FIPS ff add ORIG_FIPS according to datascale (State FIPS code)
lookupCounty$FIPS <- NULL ff remove column
smetrics <- leftJoin(smetrics, lookupCounty, by="Year")
ff Join data sets for County, PUMA, MSA, WVS Region, and State
metricsComplete<- leftJoin(metricsComplete, inmetrics[ ,c("METRIC_VAR", "FIPS", "Year", "SOURCE_ABBREV",
"metricname", "units")], by=c("METRIC_VAR", "Year", "FIPS")) ff select variables to retain
cmetrics <- cmetrics[, c("METRIC_VAR","FIPS","Year","measure","metricfipsn", "ORIG_FIPS")]
metricsComplete<- leftJoin(metricsComplete, cmetrics, by=c("METRIC_VAR", "Year", "FIPS"))
pmetrics <- pmetrics[, c("METRIC_VAR","FIPS","Year","measure","metricfipsn", "ORIG_FIPS")]
colnames(pmetrics) <- c("METRIC_VAR","FIPS","Year","measure_y","metricfipsn_y", "ORIG_FIPS_y")
metricsComplete<- leftJoin(metricsComplete, pmetrics, by=c("METRIC_VAR", "Year", "FIPS"))
metricsComplete$measure <- with(metricsComplete,
ifelse((is.na(measure) & lis.na(measure_y)), measure_y, measure))
metricsComplete$metricfipsn <- with(metricsComplete,
ifelse((is.na(metricfipsn) & lis.na(metricfipsn_y)), metricfipsn_y, metricfipsn))
metricsComplete$ORIG_FIPS <- with(metricsComplete,
ifelse((is.na(ORIG_FIPS) & !is.na(ORIG_FIPS_y)), ORIG_FIPS_y, ORIG_FIPS))
metricsComplete$measure_y <- metricsComplete$metricfipsn_y <- metricsComplete$ORIG_FIPS_y <- NULL
mmetrics <- mmetrics[, c("METRIC_VAR","FIPS","Year","measure","metricfipsn", "ORIG_FIPS")]
colnames(mmetrics) <- c("METRIC_VAR","FIPS","Year","measure_y","metricfipsn_y","ORIG_FIPS_y")
metricsComplete<- leftJoin(metricsComplete, mmetrics, by=c("METRIC_VAR", "Year", "FIPS"))
metricsComplete$measure <- with(metricsComplete,
ifelse((is.na(measure) & lis.na(measure_y)), measure_y, measure))
metricsComplete$metricfipsn <- with(metricsComplete,
ifelse((is.na(metricfipsn) & lis.na(metricfipsn_y)), metricfipsn_y, metricfipsn))
metricsComplete$ORIG_FIPS <- with(metricsComplete,
ifelse((is.na(ORIG_FIPS) & !is.na(ORIG_FIPS_y)), ORIG_FIPS_y, ORIG_FIPS))
metricsComplete$measure_y <- metricsComplete$metricfipsn_y <- metricsComplete$ORIG_FIPS_y <- NULL
Continued on next page
40
-------
Box 4.2. Continued.
rmetrics <- rmetrics[, c("METRIC_VAR","FIPS","Year","measure","metricfipsn", "ORIG_FIPS")]
colnames(rmetrics) <- c("METRIC_VAR","FIPS","Year","measure_y","metricfipsn_y", "ORIG_FIPS_y")
metricsComplete<- leftJoin(metricsComplete, rmetrics, by=c("METRIC_VAR", "Year", "FIPS"))
metricsComplete$measure <- with(metricsComplete,
ifelse((is.na(measure) & !is.na(measure_y)), measure_y, measure))
metricsComplete$metricfipsn <- with(metricsComplete,
ifelse((is.na(metricfipsn) & !is.na(metricfipsn_y)), metricfipsn_y, metricfipsn))
metricsComplete$ORIG_FIPS <- with(metricsComplete,
ifelse((is.na(ORIG_FIPS) & !is.na(ORIG_FIPS_y)), ORIG_FIPS_y, ORIG_FIPS))
metricsComplete$measure_y <- metricsComplete$metricfipsn_y <- metricsComplete$ORIG_FIPS_y <- NULL
smetrics <- smetrics[, c("METRIC_VAR","FIPS","Year","measure","metricfipsn", "ORIG_FIPS")]
colnames(smetrics) <- c("METRIC_VAR","FIPS","Year","measure_y","metricfipsn_y", "ORIG_FIPS_y")
metricsComplete<- leftJoin(metricsComplete, smetrics, by=c("METRIC_VAR", "Year", "FIPS"))
metricsComplete$measure <- with(metricsComplete,
ifelse((is.na(measure) & !is.na(measure_y)), measure_y, measure))
metricsComplete$metricfipsn <- with(metricsComplete,
ifelse((is.na(metricfipsn) & !is.na(metricfipsn_y)), metricfipsn_y, metricfipsn))
metricsComplete$ORIG_FIPS <- with(metricsComplete,
ifelse((is.na(ORIG_FIPS) & !is.na(ORIG_FIPS_y)), ORIG_FIPS_y, ORIG_FIPS))
metricsComplete$measure_y <- metricsComplete$metricfipsn_y <-metricsComplete$ORIG_FIPS_y <- NULL
metricsComplete$datafips <- metricsComplete$ORIG_FIPS
metricsComplete$datayear <- with(metricsComplete, ifelse((!is.na(measure)), Year, NA))
Table 4.5. Sample of output for finest available spatial scale by metric, county, and year. The
variable ORIG_FIPS indicates the data source, including 5-digit Federal Information Processing
Standards (FIPS) county code, 3- or 4-digit Metropolitan Statistical Area (MSA) code, 7-digit
Public Use Microdata Area (PUMA) FIPS code, 2 digit state FIPS code, or 1-digit World Values
Survey (WVS) regional code. See Appendix B for variable descriptions.
METRIC VAR
FIPS
Year
measure
metricfipsn
ORIG FIPS
datafips
datayear
ACCMM
72001
2000
3.6300
11
72
72
2000
ADLTASTHMA
72001
2000
0.2571
14
7202200
7202200
2000
ADULTLIT
72001
2000
NA
NA
NA
NA
NA
ALCOHOL
72001
2000
NA
NA
NA
NA
NA
ALLOFLFE
72001
2000
NA
NA
NA
NA
NA
ASTHMORT
72001
2000
0.0070
11
72
72
2000
41
-------
Step lc. Temporally impute missing data
Missing metric data were temporally imputed using a carry-forward substitution imputation
technique (Zhang et al., 2008). Following the methods of U.S. HWBI (Smith et al., 2014a), this
was accomplished by forward filling missing data from the previous year (Table 4.6), unless
data were not available and data were backward filled from the following year (Table 4.7), or in
combination (Table 4.8) to create a temporally complete dataset (Table 4.9) used to create fill
data for spatial imputation in processing step 2.
Box 4.3. R Code to temporally impute missing data with data from previous or following year.
domMetricsCorrected <- metricsComplete[order(metricsComplete$METRIC_VAR,
metricsComplete$FIPS, metricsComplete$Year), ] ft order by metric, FIPS, year
colnames(domMetricsCorrected)[colnames(domMetricsCorrected)== "SOURCE_ABBREV"] <- "source" ft change
column names
domMetricsCorrected <- domMetricsCorrected %>%
group_by(METRIC_VAR, FIPS) %>%
mutate(measure=na.locf(measure, na.rm=F)) %>% ft first fill forward
mutate(measure=na.locf(measure, na.rm=F, fromLast=T)) ft then fill backward
domMetricsCorrected <- domMetricsCorrected %>%
group_by(METRIC_VAR, FIPS) %>%
mutate(source=na.locf(source, na.rm=F)) %>% ft first fill forward
mutate(source=na.locf(source, na.rm=F, fromLast=T)) ft then fill backward
domMetricsCorrected <- domMetricsCorrected %>%
group_by(METRIC_VAR, FIPS) %>%
mutate(metricname=na.locf(metricname, na.rm=F)) %>% # first fill forward
mutate(metricname=na.locf(metricname, na.rm=F, fromLast=T)) ft then fill backward
domMetricsCorrected <- domMetricsCorrected %>%
group_by(METRIC_VAR, FIPS) %>%
mutate(units=na.locf(units, na.rm=F)) %>% ft first fill forward
mutate(units=na.locf(units, na.rm=F, fromLast=T)) ft then fill backward
domMetricsCorrected <- domMetricsCorrected %>%
group_by(METRIC_VAR, FIPS) %>%
mutate(datafips=na.locf(datafips, na.rm=F)) %>% ft first fill forward
mutate(datafips=na.locf(datafips, na.rm=F, fromLast=T)) ft then fill backward
domMetricsCorrected <- domMetricsCorrected %>%
group_by(METRIC_VAR, FIPS) %>%
mutate(datayear=na.locf(datayear, na.rm=F)) %>% ft first fill forward
mutate(datayear=na.locf(datayear, na.rm=F, fromLast=T)) ft then fill backward
Continued on next page
42
-------
Box 4.3. Continued.
domMetricsCorrected <- domMetricsCorrected %>%
group_by(METRIC_VAR, FIPS) %>%
mutate(metricfipsn=na.locf(metricfipsn, na.rm=F)) %>% # first fill forward
mutate(metricfipsn=na.locf(metricfipsn, na.rm=F, fromLast=T)) # then fill backward
# Estimate annual imputation error
dataError <- domMetricsCorrected! ,c("METRIC_VAR","FIPS","Year","datayear", "measure", "metricfipsn")]
dataError$measure <-with(dataError, ifelse((datayear!=Year), NA, measure))
dataError$datayear <- NULL
meanout <- dataError %>%
group_by(METRIC_VAR, FIPS) %>%
summarise(mean = mean(measure, na.rm=T)) # mean measure by metric and county without temporally imputed
(carriedforward) values
dataError <- leftJoin(dataError, meanout, by=c("METRIC_VAR", "FIPS"))
dataError$measure <- with(dataError,
ifelse((!is.na(measure)), (measure-mean), measure)) # used to calculate annual error
dataError <- dataError[!is.na(dataError$metricfipsn) & dataError$metricfipsn>2, c("METRIC_VAR", "Year", "FIPS",
"measure") ]
dataError$MetricFIPS <- paste(dataError$METRIC_VAR, dataError$FIPS, sep="_")
dataError <- dataError[,c("MetricFIPS", "Year", "measure")]
dataError <-spread(dataError, MetricFIPS, measure)
dataError <- data.frame(dataError, check.names=F)
df <- data.frame(dataError$Year)
colnames(df) <- "Year"
# Function used to iteratively calculate error across years
for(i in 2:ncol(dataError)) {
model <- lm(dataError[,i] ~ dataError$Year)
z <- dataError[,i]
z <- data.frame(z, check.names=F)
annualerror <- predict.lm(model, z, se.fit=T, interval="confidence")$se.fit
df <- cbind(df,annualerror)
}
colnames(df) <- colnames(dataError)
annualerror <-gather(df, "MetricFIPS", "annualerror", 2:ncol(df))
annualerror <- annualerror %>% separate(MetricFIPS, into = c("METRIC_VAR","FIPS"), sep = "_")
annualerror$FIPS <- as.numeric(annualerror$FIPS)
domMetricsCorrected <- leftJoin(domMetricsCorrected, annualerror,
by=c("METRIC_VAR", "Year", "FIPS"))
domMetricsCorrected$annualerror <- with(domMetricsCorrected,
ifelse(annualerror==0, NA,
ifelse((Year==datayear), 0, annualerror))) ft step 1 output
43
-------
Table 4.6. Example of temporal imputation where missing metric data (measure) were
forward filled with data from the previous year. In this example, missing metrics (where
ORIG_FIPS=NA) for 2008 and 2009 were filled with data from 2007. See Appendix B for variable
descriptions.
METRICVAR
FIPS
Year
measure
metricfipsn
ORIGFIPS
datafips
datayear
ASTHMORT
72001
2000
0.0070
11
72
72
2000
ASTHMORT
72001
2001
0.0048
11
72
72
2001
ASTHMORT
72001
2002
0.0021
11
72
72
2002
ASTHMORT
72001
2003
0.0042
11
72
72
2003
ASTHMORT
72001
2004
0.0043
11
72
72
2004
ASTHMORT
72001
2005
0.0043
11
72
72
2005
ASTHMORT
72001
2006
0.0037
11
72
72
2006
ASTHMORT
72001
2007
0.0036
11
72
72
2007
ASTHMORT
72001
2008
0.0036
11
NA
72
2007
ASTHMORT
72001
2009
0.0036
11
NA
72
2007
ASTHMORT
72001
2010
0.0020
11
72
72
2010
Table 4.7. Example of temporal imputation where missing metric data (measure) were
backward filled with data from the following year. In this example, missing metrics (where
ORIG_FIPS=NA) for 2000-2004 were filled with data from 2005. See Appendix A for description
of variable names.
METRICVAR
FIPS
Year
measure
metricfipsn
ORIGFIPS
datafips
datayear
UNIVGRAD
72001
2000
0.1141
9
NA
7202200
2005
UNIVGRAD
72001
2001
0.1141
9
NA
7202200
2005
UNIVGRAD
72001
2002
0.1141
9
NA
7202200
2005
UNIVGRAD
72001
2003
0.1141
9
NA
7202200
2005
UNIVGRAD
72001
2004
0.1141
9
NA
7202200
2005
UNIVGRAD
72001
2005
0.1141
9
7202200
7202200
2005
UNIVGRAD
72001
2006
0.1403
9
7202200
7202200
2006
UNIVGRAD
72001
2007
0.1236
9
7202200
7202200
2007
UNIVGRAD
72001
2008
0.1566
9
7202200
7202200
2008
UNIVGRAD
72001
2009
0.1482
9
7202200
7202200
2009
UNIVGRAD
72001
2010
0.1155
9
7202200
7202200
2010
44
-------
Table 4.8. Example of temporal imputation where missing metric data (measure) were both
forward and backward filled with data from previous and following years. In this example,
missing metrics (where ORIG_FIPS=NA) for 2000 and 2002-2010 were filled with data from
2001. See Appendix B for variable descriptions.
METRICVAR
FIPS
Year
measure
metricfipsn ORIGFIPS
datafips
datayear
HAPPY
72001
2000
0.8889
1 NA
3
2001
HAPPY
72001
2001
0.8889
1 3
3
2001
HAPPY
72001
2002
0.8889
1 NA
3
2001
HAPPY
72001
2003
0.8889
1 NA
3
2001
HAPPY
72001
2004
0.8889
1 NA
3
2001
HAPPY
72001
2005
0.8889
1 NA
3
2001
HAPPY
72001
2006
0.8889
1 NA
3
2001
HAPPY
72001
2007
0.8889
1 NA
3
2001
HAPPY
72001
2008
0.8889
1 NA
3
2001
HAPPY
72001
2009
0.8889
1 NA
3
2001
HAPPY
72001
2010
0.8889
1 NA
3
2001
Table 4.9. Sample step 1 output where original data source is identified (ORIG_FIPS) and
missing data have been temporally imputed. See Appendix B for variable descriptions.
METRICVAR
FIPS
Year
measure
metricfipsn
ORIGFIPS
datafips
datayear
annualerror
ACCMM
72001
2000
3.63
11
72
72
2000
0
ACCMM
72001
2001
3.78
11
72
72
2001
0
ACCMM
72001
2002
3.44
11
72
72
2002
0
ACCMM
72001
2003
2.95
11
72
72
2003
0
ACCMM
72001
2004
2.82
11
72
72
2004
0
ACCMM
72001
2005
3.16
11
72
72
2005
0
Processing step 2: Create fill data for spatial imputation
A mean value imputation method was used as a substitute for missing county-level metric data
points when metric data were missing or not based on a municipio-specific spatial scale. Prior
to imputation, metric data were summarized based on a combination of the United States
Department of Agriculture (USDA) Rural-Urban Continuum Code (RUCC) classifications (United
States Department of Agrigulture, 2013) and the GINI Index (GINI) for household income
inequality quintile bandings (United States Census Bureau, 2012b). These combinations
reflected similarities in well-being based on spatial relationships with large urban centers and
the dispersion of income (Smith et al., 2014a). RUCC is indexed 1-9 (Metropolitan -
Nonmetropolitan) and GINI is indexed 0-1 (perfect equality-perfect inequality). For imputation,
GINI were condensed into 5 categories (1-5) based on quintile distribution. Fill data grouped by
45
-------
metric, year, and RUCC-GINI combination (Table 4.10) were used for spatial imputation in
processing step 3.
Box 4.4. R Code to group data by metric, year, and RUCC-GINI combination in preparation for
spatial imputation.
ft Determine GINI index quintile distribution for U.S. and apply to Puerto Rico
xOO <- domains_metrics[domains_metrics$Year=="2000" & domains_metrics$METRIC_VAR=="ACCMM", ] ftsubset
so there is only 1 entry per county
lookupUSA <- leftJoin('xOO//c("lFIPS", "STATE_ABBR", "gini_idx_band"//, x, by="FIPS"/) ft build lookup table with GINI
index value by U.S. county (n=3,143)
giniBands <- as.matrix(quar\t\\e(lookupUSA$GINI_INDEX, probs=seq(0,l,0.2))) ft determine 20% quintile distribution
of U.S. GINI index value
gini <- read.csv("'./gini_indices.csv"/ header=T/ stringsAsFactors=F/) ft file containing all U.S. and territory GINI index
giniPR <- gini[gini$FIPS>=72000, ] ft subset Puerto Rico by municipio (n=78)
giniPR$GINI_IDX_BAND <- with (giniPR,
\fe\se(GINI_INDEX>=giniBands[l,l] & GINI_INDEX=giniBands[2,l] & GINI_INDEX=giniBands[3,l] & GINIJNDEX=giniBands[4,l] & GINI_INDEX=giniBands[5/l] & GINI_INDEX<=giniBands[6/l])/ 5, NAJJJJJJ ft apply U.S. quintiles (1 -5) to
Puerto Rico GINI index values
ft Join RUCC-GINI identifiers to metric data for imputations
fipsCounty <- lookupFIPS[lookupFIPS$SCALE=="County", c("FIPS","Year","RUCC","GINI_IDX_BAND")] ft merge in
RUCC, gini_idx_band
domMetricsCorrected <- leftJoin(domMetricsCorrected, fipsCounty, by=c("FIPS", "Year"))
ft Create fill data (Puerto Rico) for spatial imputation
lookupCounty <- lookupFIPS[lookupFIPS$SCALE=="County", c("FIPS7,Year7,PUMA_REGI0N","MSA_REGI0N",
"WVS_REGION","STATE_ABBR")] ft Join geographic attributes to domMetricsCorrected
domMetricsCorrected <- leftJoin(domMetricsCorrected, lookupCounty, by=c("FIPS", "Year"))
ft Summarize data grouped by RUCC-GINI combination
RGmed <- domMetricsCorrected %>%
group_by(METRIC_VAR, Year, RUCC, GINI_IDX_BAND) %>%
summarise(RG_MED = median(measure, na.rm=T)) ft median by metric, year, RUCC, and GINI
RGmad <- domMetricsCorrected %>%
group_by(METRIC_VAR, Year, RUCC, GINI_IDX_BAND) %>%
summarise(RG_MAD = mad(measure, constant=l, na.rm=T)) ft median absolute difference about the median by
metric, year, RUCC, and GINI
RGmean <- domMetricsCorrected %>%
group_by(METRIC_VAR, Year, RUCC, GINI_IDX_BAND) %>%
summarise(RG_MEAN = mean(measure, na.rm=T)) ft mean by metric, year, RUCC, and GINI
Continued on next page
46
-------
Box 4.4. Continued.
RGerr <- domMetricsCorrected %>%
group_by(METRIC_VAR, Year, RUCC, GINI_IDX_BAND) %>%
summarise(RG_ERR = sqrt(var(measure,na.rm=TRUE)/length(na.omit(measure)))) # standard error of the mean by
metric, year, RUCC, and GINI
RGobs <- domMetricsCorrected %>%
group_by(METRIC_VAR, Year, RUCC, GINI_IDX_BAND) %>%
summarise(RG_OBS = length(measure)) # sample size (n) by metric, year, RUCC, and GINI
ruccginiData <- leftJoin(RGobs, RGmean, by=c("METRIC_VAR", "Year", "RUCC", "GINI_IDX_BAND"))
ruccginiData <- leftJoin(ruccginiData, RGerr, by=c("METRIC_VAR", "Year", "RUCC", "GINI_IDX_BAND"))
ruccginiData <- leftJoin(ruccginiData, RGmed, by=c("METRIC_VAR", "Year", "RUCC", "GINI_IDX_BAND"))
ruccginiData <- leftJoin(ruccginiData, RGmad, by=c("METRIC_VAR", "Year", "RUCC", "GINI_IDX_BAND"))
ruccginiData <- ruccginiData[!is.na(ruccginiData$RG_MEAN), ] # remove unnecessary RUCC-GINI combinations
# Create final fill data for spatial imputation
domMetricsFillRG <- ruccginiData
colnames(domMetricsFillRG) <- c("METRIC_VAR","Year","RUCC","GINI JDX_BAND","FILL_OBS","FILL_MEAN",
"FILL_ERR","FILL_MED","FILL_MAD")
domMetricsFillRG$FILL_SOURCE <- "RUCC-GINI" #step 2 output
Table 4.10. Sample step 2 output (Domain Metrics Fill (RUCC-GINI) used for spatial
imputation). See Appendix B for variable descriptions.
FILL_
FILL_
FILL_
FILL_
FILL_
FILL_
METRICVAR
Year
RUCC
GINIIDXBAND
OBS
MEAN
ERR
MED
MAD
SOURCE
ACCMM
2000
1
4
3
4.230
0.000
4.23
0
RUCC-GINI
ACCMM
2000
1
5
37
4.147
0.047
4.23
0
RUCC-GINI
ACCMM
2000
2
5
16
3.638
0.033
3.63
0
RUCC-GINI
ACCMM
2000
3
3
1
3.220
NA
3.22
0
RUCC-GINI
ACCMM
2000
3
5
12
3.493
0.058
3.63
0
RUCC-GINI
ACCMM
2000
4
5
3
3.630
0.000
3.63
0
RUCC-GINI
Processing step 3: Complete metric data with spatial imputations
Metric data that were either missing or not based on county-level data were spatially
imputation with the median metric value by RUCC-GINI combination (Figure 4.2). For metrics
that were available for Puerto Rico but based on a spatial area coarser than county-level, data
were summarized by RUCC-GINI combination and used for imputation. Metrics that were not
available for Puerto Rico at any spatial scale or based on units of measure that could not be
standardized to U.S. data were removed from HWBI unless removal resulted in a missing
indicator, in which case U.S. data summarized by RUCC-GINI combination were used for
47
-------
imputation. Spatially complete metric data (Table 4.11) were used for normalization of
standardization in processing step 4.
No-
Yes-
-Yes-
¦No-
No-
Yes
-Yes-
No-
Imputed from U.S.
by Rucc-Gini
Metric removed
Imputed from Puerto
Rico by Rucc-Gini
Unimputed
Are data available for
any higher-level
geography?
Is metric comparable
to U.S. HWBI (not
substitute measure
and units)?
Are data available by
municipio (after temporal
imputation)?
Will removing metric
result in the loss of an
indicator?
Figure 4.2. Summary of spatial imputation of metrics for Puerto Rico HWBI.
Box 4.5. R Code to impute missing data based on RUCC-GINI code.
# Update county-level metric template to include all U.S. metrics, as some missing metrics are necessary to avoid
missing indicators
metricsComplete <- merge(lookupMetrics[ ,c("DOMAIN","INDICATOR","METRIC_VAR","POS_NEG_METRIC", "units",
"PR_modification")], uniqueFips) # get complete metric x county x year template for all U.S. metrics
domMetricsCorrectedAII <- left_join(metricsComplete,
select(domMetricsCorrected, -c(DOMAIN, INDICATOR, POS_NEG_METRIC,
RUCC, GINI_IDX_BAND, source, units, PUMA_REGION, MSA_REGION,
WVS_REGION, STATE_ABBR)), by=c("METRIC_VAR", "FIPS", "Year"))
ft Join geographic attributes to dom.metrics_corrected
lookupCounty <- lookupFIPS[lookupFIPS$SCALE=="County", c("FIPS","Year","PUMA_REGION","MSA_REGION",
"WVS_REGION","STATE_ABBR","RUCC","GINI_IDX_BAND")]
domMetricsCorrectedAII <- leftJoin(domMetricsCorrectedAII, lookupCounty, by=c("FlPS", "Year"))
Continued on next page
48
-------
Box 4.5. Continued.
ft Identify metrics as unimputed or not
inmetrics <- select(domMetricsCorrectedAII, -c(metricfipsn, ORIG_FIPS))
inmetrics$unimputed <- with(inmetrics, ifelse((datafips==FIPS & lis.na(datafips)), "Y", "N"))
ft Create lookup table of Puerto Rico fill data
fillPR <- select(domMetricsFillRG, -c(FILL_ERR, FlLL_MEAN))
fillPR$Year <- as.numeric(fillPR$Year)
fillPR$RUCC <- as.numeric(fillPR$RUCC)
fillPR$GINl_lDX_BAND <- as.numeric(fillPR$GINI_IDX_BAND)
fillPR$METRIC VAR <- as.character(fillPR$METRIC_VAR)
ft Create lookup table of U.S. fill data
fillUS <- USdomMetricsFillRG
fillUS$FILL_SOURCE <- with(fillUS, ifelse(FILL_SOURCE=="RUCC-GINI", "RUCC-GINIJJS",
ifelse((FILL_SOURCE=="RUCC ONLY"), "RUCC ONLY_US", NA)))
fillUS <- select(fillUS, -c(FILL_ERR, FILL_MEAN))
colnames(fillUS) <- c("FILL_SOURCE_US","METRIC_VAR","Year","RUCC", "GlNl_lDX_BAND"/'FILL_MED_US",
"FILL_MAD_US","FILL_QSTD_US","FILL_OBS_US")
ft Merge metric data, Puerto Rico fill data, and U.S. fill data for imputation
metrics <- left_join(inmetrics, fillPR, by=c("METRIC_VAR", "Year", "RUCC", "GINl_IDX_BAND"))
metrics <- leftJoin(metrics, fillUS, by=c("METRIC_VAR", "Year", "RUCC", "GINI_IDX_BAND"))
ft Imputation
metrics$measure <- with(metrics,
ifelse(unimputedl="Y" & ft if unimputed is not Y,
!is.na(PR_modification) & ft and data were not missing for Puerto Rico,
PR_modification!="Substitute measure and units" & ft and metric was not substitute measure and units,
METRIC_VAR!="LEISURE", ft and metric was not LEISURE (LEISURE was derived from Nicaragua and Buenos Aires,
therefore for U.S. comparison we are imputing from U.S. RUCC-GINI)
FILL_MED, ft then impute from Puerto Rico RUCC-Gini
ifelse(unimputed!="Y" & # if unimputed is not Y,
is.na(PR_modification) | # and data were missing for Puerto Rico,
PR_modification=="Substitute measure and units", # or metric was substitute measure and units,
FILL_MED_US, # then impute from U.S. RUCC-Gini
ifelse((METRIC_VAR=="LEISURE"), # and metric was LEISURE (LEISURE was derived from Nicaragua and Buenos
Aires, therefore for U.S. comparison we are imputing from U.S. RUCC-GINI),
FILL_MED_US, measure)))) ft then impute from U.S. RUCC-Gini
metrics$FILL_SOURCE <- with(metrics, ifelse(unimputed!="Y" &
is.na(PR_modification) | PR_modification=="Substitute measure and units", FILL_SOURCE_US,
ifelse((METRIC_VAR=="LEISURE"), FILL_SOURCE_US, FILL_SOURCE)))
metrics$FILL_MED <- with(metrics, ifelse(unimputed!="Y" &
is.na(PR_modification) | PR_modification=="Substitute measure and units", FILL_MED_US,
ifelse((METRIC_VAR=="LEISURE"),FILL_MED_US, FILL_MED)))
Continued on next page
49
-------
Box 4.5. Continued.
metrics$FILL_MAD <- with(metrics, ifelse(unimputed!="Y" &
is.na(PR_modification) | PR_modification=="Substitute measure and units", FILL_MAD_US,
ifelse((METRIC_VAR=="LEISURE"),FILL_MAD_US, FILL_MAD)))
metrics$FILL_OBS <- with(metrics, ifelse(unimputed!="Y" &
is.na(PR_modification) | PR_modification=="Substitute measure and units", FILL_OBS_US,
ifelse((METRIC_VAR=="LEISURE"),FILL_OBS_US, FILL_OBS)))
metrics$FILL_OBS_US <- metrics$FILL_MED_US <- metrics$FILL_MAD_US <- metrics$FILL_SOURCE_US <- NULL
# Determine original data scale (finest available spatial scale) unless imputed by U.S.
metrics$datascale <- with(metrics,
ifelse(FILL_SOURCE=="RUCC-GINI_US", "U.S.",
ifelse(FILL_SOURCE=="RUCC ONLY_US", "U.S.",
ifelse(datafips==0, "National",
ifelse(datafips>7200000, "PUMA Region",
ifelse(datafips>=100 & datafips<=1400, "MSA Region",
ifelse(datafips>=l & datafips<=6, "WVS Region",
ifelse((datafips==72), "State", "County"))))))))
domMetrics <- metrics[!is.na(metrics$measure), ] ft step 3 output
50
-------
Table 4.11. Sample step 3 output (metrics completed with spatial imputation). See Appendix B for variable descriptions.
_o
-Z.
l/l
cc
ru
_u
w—
'~o
o
E
1
en
<
LU
Z5
METRICVA
(-0
Q_
ru
QJ
measure
CO
Q.
M—
ru
ru
i_
ru
QJ
>•
ru
ru
nnualerror
u
u
Z)
CD
1
X
n
_i
nimputed
ILL_OBS
a
L_U
1
n
<
i
u
a:
Z)
o
LO
i
1
Q
1—
8
i
Q_
LL
>
~o
"O
ru
a:
=3
LL
LL
LL
Ll_
Ll_
BEAUSPRT
NA
72001
2000
0.79
NA
NA
NA
6
5
N
123
0.789
0.023
RUCC-GINIUS
0.017
U.S.
ALLOFLFE
Substitute
measure
72001
2000
0.97
3
2001
NA
6
5
N
4
0.970
0.000
RUCC-GINI
0.023
WVS
Region
TOTRATE
Modified
source
Substitute
72001
2000
933.3
3
2001
NA
6
5
N
4
933.3
0.000
RUCC-GINI
134.689
WVS
Region
PERARTS
measure
and units
72001
2000
0.45
72001
2013
NA
6
5
Y
123
0.454
0.075
RUCC-GINIUS
0.112
U.S.
MATHTEST
Modified
source
72001
2000
0.70
72001
2009
NA
6
5
Y
4
0.748
0.017
RUCC-GINI
0.117
County
READIEST
Modified
source
72001
2000
0.81
72001
2009
NA
6
5
Y
4
0.815
0.007
RUCC-GINI
0.057
County
51
-------
Processing step 4: Normalize and standardize metric values
To create a composite index comprised of metric with different units of measure and
distributions, the U.S. HWBI followed the normalization and standardization procedure used by
the Organisation for Economic Co-operation and Development Better Life Index (Organisation
for Economic Co-operation and Development, 2011).
Step 4a. Normalize metric values
Minimum and maximum values were identified for each metric as three interquartile ranges
less than or greater than the first or third quartiles, respectively. Metric values beyond the
minimum or maximum were identified as outliers.
Box 4.6. R Code to normalize metric values.
metrics <- domMetrics[domMetrics$Year<=2010, ] #subset to include only mean decadal (2000-2010)
# Remove metrics that are completely missing for Puerto Rico, unless that is the single metric within an indicator
metrics <- metrics[!is.na(metrics$PR_modification) & # keep metrics that were not missing for Puerto Rico
metrics$METRIC_VAR!="PERARTS" & metrics$METRIC_VAR!="VACATION", ] # remove metrics that were
substitute measure and units, unless removal resulted in the elimination of an indicator
metrics <- leftJoin(metrics, meanOutUS, by="METRIC_VAR") ft join U.S. interguartile ranges to normalize
Puerto Rico metric values
metrics$fencel <- metrics$ql-(metrics$qrange*3) # identify three interguartile ranges less than first guartile
metrics$fence2 <- metrics$q3+(metrics$qrange*3) # identify three interguartile ranges greater than third guartile
metrics$outlier <- with(metrics, ifelse((measure>=fencel & measure<=fence2), "0", "1")) # identify outliers
Step 4b. Standardize metric values
Normalized metric values were standardized using OECD's formula which converts the original
values of the metrics into proportions that range between 0 (for the worst possible outcome)
and 1 (for the best possible outcome). However this scale was modified for HWBI and data
were scaled 0.1 - 0.9 to allow for potential improvements and declines beyond what was
observed in the data. Metrics reflecting positive or negative well-being were corrected in this
step to ensure that values approaching one reflected greater well-being (e.g., diabetes
mortality as a percentage of total deaths [DIABMORT] was corrected so lower mortality
reflected greater well-being). Temporally and spatially complete metric data that were
normalized and standardized (Table 4.12) were used for HWBI calculation in Chapter 5.
52
-------
Box 4.7. R Code to standardize metric values between 0.1 and 0.9.
ft Standardize positive and negative metrics to 0.1 - 0.9, where metric values reflecting less well-being are closer to 0
and greater well-being are closer to 1
metrics$NUMERATOR <- metrics$measure-metrics$MINVAL
metrics$DENOMINATOR <- metrics$MAXVAL-metrics$MINVAL
metrics$ORIG_UNIT <- metrics$units
metrics$METRIC_VAL <- with(metrics,
ifelse(!is.na(DENOMINATOR) & DENOMINATORS & POS_NEG_METRIC=="P",
(.9-.l)*NUMERATOR/DENOMINATOR+0.1,
ifelse((!is.na(DENOMINATOR) & DENOMINATOR>0 & POS_NEG_METRIC=="N"),
l-((.9-.l)*NUMERATOR/DENOMINATOR+.l), NA)))
# Set outlier observation to 0.1 or 0.9
metrics$METRIC_VAL <- with(metrics, ifelse(METRIC_VAL<0.1, 0.1, ifelse((METRIC_VAL>0.9), 0.9, METRIC_VAL)))
# Correct units of standardized metrics, now all proportions 0.1 - 0.9
metrics$units <- with(metrics,
ifelse(!is.na(DENOMINATOR) & DENOMINATORS & POS_NEG_METRIC=="P","Proportion",
ifelse((!is.na(DENOMINATOR) & DENOMINATOR>0 & POS_NEG_METRIC=="N"),"Proportion", NA)))
# Remove any missing values
domMetricsStandard <- metrics[!is.na(metrics$METRIC_VAL), ] ft step 4 output
53
-------
Table 4.12. Sample step 4 output (metrics normalized and standardized). See Appendix B for variable descriptions.
u
Row
DOMAIN
INDICATOR
METRICVAR
POS_NEG_METRI
units
PRmodification
FIPS
Year
measure
datafips
datayear
annualerror
RUCC
GINIIDXBAND
unimputed
FILL_OBS
Q
LU
I
1
1
LL
1
Connection
to Nature
Biophilia
ALLOFLFE
P
Proportion
Substitute
measure
72001
2000
0.97
3
2001
NA
6
5
N
4
0.97
2
Cultural
Fulfillment
Cultural Activity
Participation
Basic
TOTRATE
P
Proportion
Modified
source
72001
2000
933.3
3
2001
NA
6
5
N
4
933.3
3
Education
Educational
Knowledge and
Skills of Youth
Basic
MATHTEST
P
Proportion
Modified
source
72001
2000
0.7
72001
2009
NA
6
5
Y
4
0.75
4
Education
Educational
Knowledge and
Skills of Youth
Basic
READIEST
P
Proportion
Modified
source
72001
2000
0.81
72001
2009
NA
6
5
Y
4
0.82
5
Education
Educational
Knowledge and
Skills of Youth
SCITEST
P
Proportion
Modified
source
72001
2000
0.76
72001
2009
NA
6
5
Y
4
0.82
6
Education
Participation
and Attainment
ADULTLIT
N
Proportion
Modified
source
72001
2000
0.09
72
2004
NA
6
5
N
4
0.09
Columns continue on next page
54
-------
Table 4.12 Columns continued.
Row
FILL_MAD
FILL_SOURCE
FILL_QSTD_US
datascale
t—1
cr
m
cr
qrange
outlier
MINVAL
MAXVAL
c
nALL
scalefactor
impactfactor
NUMERATOR
DENOMINATOR
ORIGUNIT
METRIC_VAL
1
0
RUCC
-GINI
0.02
WVS
Region
0.64
0.65
0
1
0.64
0.65
27786
34573
0.14
0.2
0.33
0.01
Percentage
0.9
2
0
RUCC
-GINI
134.7
WVS
Region
391.8
644.0
252.2
0
18.2
1398.0
34549
34573
0.72
0
915.2
1379.8
Rate
0.63
3
0.02
RUCC
-GINI
0.12
County
0.7
0.78
0.08
0
0.56
0.83
34573
34573
1
0
0.14
0.27
Percentage
0.51
4
0.01
RUCC
-GINI
0.06
County
0.68
0.72
0.04
0
0.65
0.76
34573
34573
1
0
0.17
0.12
Percentage
0.9
5
0.03
RUCC
-GINI
0.04
County
0.66
0.75
0.09
0
0.63
0.77
34573
34573
1
0
0.13
0.14
Percentage
0.82
6
0
RUCC
-GINI
0.03
State
0.09
0.14
0.05
0
0.08
0.17
34573
34573
1
0
0.02
0.09
Percentage
0.75
55
-------
Chapter 5: Calculating HWBI
5.1. '
-------
chosen, rather than 0 to 1, to allow for the possibility of declines or improvements beyond what
was observed in the data. Note a calculated HWBI for a given year or location will likely still be
well above the minimum possible HWBI (0.1) or well below the maximum possible HWBI (0.9),
because calculated values likely reflect a mixture of low, medium, and high scoring metrics.
Therefore it is useful to compare a given score relative to calculated comparables (e.g., other
states, counties, or years).
The HWBI can be customized to a specific time period or spatial scale according to user needs
using population to weight metrics during indicator score calculation. Additionally, community-
specific values can be incorporated into the index. For the U.S. HWBI, Relative Importance
Values (RIVs) ranked the contribution of each domain to overall well-being (Smith et al., 2013a).
These RIVs were based on theoretical relationships between human well-being and ecosystem
services, expert opinion from an interdisciplinary panel, and a convenience survey proxy for
public opinion. As such, they were intended as a conceptual place-holder until more interactive,
community-specific tools were developed. Since no Puerto Rico-specific RIVs have been
determined, the eight domains of human well-being were equally weighted in all U.S. and
Puerto Rico HWBI calculations within this report.
5.2. Calculating HWBI
The input data for calculating HWBI is the standardized metric data for all municipios across all
available years, where each row is a metric x county x year combination (Table 4.12). In this
example we calculate the mean decadal (2000-2010) HWBI by municipio (county equivalent).
Processing step 1: Mean decadal (2000 - 2010) indicator score calculation by municipio
Indicator scores are the arithmetic means of standardized metrics by indicator. In this example
HWBI was calculated by municipio for 2000-2010 decade, therefore annual population was
used to weight standardized metric values. Output was organized by municipio (unique Federal
Information Processing Standards [FIPS] code) and indicator (Table 5.1).
Box 5.1. R Code to calculate indicator scores as arithmetic mean of metrics, by municipio and
weighted population across years.
x <- left_join(domMetricsStandard, lookupFIPS[lookupFIPS$SCALE=="County", cf'FIPS", "Year", "POPULATION")],
by=c("FIPS", "Year"))
indicatorScore <- x %>%
group_by(FIPS, INDICATOR) %>%
summarise(lndicatorScore = weighted.mean(METRIC_VAL, POPULATION)) # arithmetic mean by indicator and
county and weighted by population across years
57
-------
Table 5.1. Sample output of Puerto Rico mean decadal (2000-2010) HWBI indicator scores by
municipio (unique Federal Information Processing Standards [FIPS] code).
FIPS INDICATOR IndicatorScore
72001 Actual Safety 0.79
72001 Attitude toward Others and the Community 0.39
72001 Basic Educational Knowledge and Skills of Youth 0.73
72001 Basic Necessities 0.36
72001 Biophilia 0.90
72001 Cultural Activity Participation 0.63
Processing step 2: Mean decadal (2000 - 2010) domain score calculation by municipio
Domain scores are arithmetic means of indicator scores by domain. Output was organized by
municipio and domain (Table 5.2).
Box 5.2. R Code to calculated domain scores as arithmetic means of indicator scores by
municipio.
# Create indicator-domain lookup table
indDom <-do.call(rbind, by(lookupMetrics, list(lookupMetrics$INDICATOR), FUN=function(x) head(x, 1)))
indDom <- leftJoin(indicatorScore, indDom[,c("DOMAIN", "INDICATOR")], by="INDICATOR")
domainScore <- indDom %>%
group_by(FIPS, DOMAIN) %>%
summarise(DomainScore = mean(lndicatorScore)) # arithmetic mean by domain and county
Table 5.2. Sample output of Puerto Rico mean decadal (2000-2010) HWBI domain scores by
municipio (unique Federal Information Processing Standards [FIPS] code).
FIPS
DOMAIN
DomainScore
72001
Connection to Nature
0.90
72001
Cultural Fulfillment
0.63
72001
Education
0.42
72001
Health
0.53
72001
Leisure Time
0.44
72001
Living Standards
0.28
72001
Safety and Security
0.49
72001
Social Cohesion
0.42
58
-------
Processing step 3: Mean decadal (2000 - 2010) HWBI calculation by municipio
The composite index value of the HWBI is the geometric mean of the domain scores. For
Puerto Rico, the domain contributions to overall well-being were not weighted by RIV,
therefore all eight domains were weighted equally. Output was organized by municipio (Table
5.3).
Box 5.3. R Code to calculated composite HWBI as geometric mean of domain scores for each
municipio.
HWBI <- domainScore %>%
group_by(FIPS) %>%
summarise(HWBI = prod(DomainScore)A(l/length(DomainScore))) # geometric mean by county
Table 5.3. Sample output of Puerto Rico mean decadal (2000-2010) HWBI by municipio
(unique Federal Information Processing Standards [FIPS] code).
FIPS HWBI
72001 049
72003 0.51
72005 0.51
72007 0.51
72009 0.51
72011 0.51
5.3. Spatial and Temporal Flexibility in Calculation
The HWBI is flexible in its spatial and temporal resolution depending on user needs. By
incorporating population weights across time and/or space, the HWBI can be calculated for a
specific time period and/or area. In the HWBI calculation above, the mean decadal HWBI by
municipio was based on across-year population weighted indicator scores (Table 5.3). However
modifying the index to a county-level annual time series was unweighted because metric data
were collected annually by municipio (Table 5.4). In another example, the mean decadal state-
level HWBI was weighted based on both across-county and across-year populations (Table 5.5).
59
-------
Example 1: Annual HWBI calculation by municipio
Box 5.4. R Code to calculate HWBI annually for each municipio.
x <- left_join(domMetricsStandard, lookupFIPS[lookupFIPS$SCALE=="County", cf'FIPS", "Year", "POPULATION")],
by=c("FIPS", "Year"))
# Calculate indicator scores
indicatorScore <- x %>%
group_by(FIPS, Year, INDICATOR) %>%
summarise(lndicatorScore = weighted.mean(METRIC_VAL)) # arithmetic mean by indicator, county, and year
(unweighted)
# Create indicator-domain lookup table
indDom <-do.call(rbind, by(lookupMetrics, list(lookupMetrics$INDICATOR), FUN=function(x) head(x, 1)))
indDom <- leftJoin(indicatorScore, indDom[,c("DOMAIN", "INDICATOR")], by="INDICATOR")
# Calculate domain scores
domainScore <- indDom %>%
group_by(FIPS, Year, DOMAIN) %>%
summarise(DomainScore = mean(lndicatorScore)) # arithmetic mean by domain, county, and year
# Calculate Human Well-being Index
HWBI <- domainScore %>%
group_by(FIPS, Year) %>%
summarise(HWBI = prod(DomainScore)A(l/length(DomainScore))) # geometric mean by county and year
Table 5.4 Sample of Puerto Rico HWBI by year.
FIPS
Year
HWBI
72001
2000
0.48
72001
2001
0.48
72001
2002
0.48
72001
2003
0.47
72001
2004
0.48
72001
2005
0.47
60
-------
Example 2: Mean decadal (2000-2010) HWBI calculation by commonwealth (state equivalent)
Box 5.5. R Code to calculate mean decadal HWBI by "state" or equivalent (i.e., Puerto Rico).
domMetricsStandard <- left_join(domMetricsStandard,
lookupFIPS[lookupFIPS$SCALE=="County" & lookupFIPS$Year<=2010, cf'FIPS", "Year", "POPULATION")],
by=c("FIPS", "Year"))
# Calculate indicator scores
indicatorScore <- domMetricsStandard %>%
group_by(STATE_ABBR, INDICATOR) %>%
summarise(lndicatorScore = weighted.mean(METRIC_VAL, POPULATION)) # arithmetic mean by Indicator and
state, weighted by population
# Create indicator-domain lookup table
indDom <-do.call(rbind, by(lookupMetrics, list(lookupMetrics$INDICATOR), FUN=function(x) head(x, 1)))
indDom <- left_join(indicatorScore, indDom[,c("DOMAIN", "INDICATOR")], by="INDICATOR")
# Calculate domain scores
domainScore <- indDom %>%
group_by(STATE_ABBR, DOMAIN) %>%
summarise(DomainScore = mean(lndicatorScore)) # arithmetic mean by domain and state (indicator scores were w
eighted by population)
# Calculate Human Well-being Index
HWBI <- domainScore %>%
group_by(STATE_ABBR) %>%
summarise(HWBI = prod(DomainScore)A(l/length(DomainScore))) # geometric mean by state (indicator scores
were weighted by population)
Table 5.5. Puerto Rico mean decadal (2000-2010) HWBI.
STATE ABBR
HWBI
PR
0.51
61
-------
Chapter 6: Mapping HWBI
6.1. Mapping HWBI in R
Mapping the HWBI allows for the visual assessment of human well-being and comparison
among regions. Here we demonstrate mapping procedures using Geographic Information
System (GIS) functionality using publically and freely available packages within R.
All mapping was performed using R version 3.1.2 using the packages rgdal version 1.0.6 (Bivand
et al., 2015) spatial data processing and ggplot2 version 1.0.1(Wickham and Chang, 2015) for
plotting maps.
A shapefile of Puerto Rico municipios was processed by extracting polygons as a data frame and
joining shapefile attributes (Table 6.1). The data were mapped using either a simple base R
function (Figure 6.1) or customized using the package ggplot2 for more customized mapping
(Figures 6.2-6.4).
Box 6.1. R Code to merge attributes with polygons in shapefile.
municipios <- readOGR(dsn = "./PR_municipios", "PR municipios proj clip")
## OGR data source with driver: ESRI Shapefile
## Source: "./PR_municipios", layer: "PR_municipios_proj_clip"
## with 78 features
## It has 17 fields
municipios. f <- fortify(municipios, region-CNTYIDFP') # fortify extracts polygons as a data frame
municipios.fm <- merge(municipios.f, municipios@data, by.x="id", by.y="CNTYIDFP") #merge attributes with
polygons in fortified data frame
plot(municipios, border="darkgray") #simple plot
Table 6.1. Sample of fortified data frame of polygons joined with shapefile attributes. Input
shapefile layer includes variables identifying longitude, latitude, and identifiers for polygons of
Puerto Rico municipios.
id
long
lat
order
hole
piece
group
STATEFP
COUNTYFP
COUNTYNS
NAME
72001
746449.4
2008656
1
FALSE
1
72001
1
72
001
01804480
Adjuntas
72001
746444.0
2008591
2
FALSE
1
72001
1
72
001
01804480
Adjuntas
72001
746442.5
2008577
3
FALSE
1
72001
1
72
001
01804480
Adjuntas
72001
746444.6
2008563
4
FALSE
1
72001
1
72
001
01804480
Adjuntas
72001
746458.1
2008481
5
FALSE
1
72001
1
72
001
01804480
Adjuntas
72001
746457.5
2008470
6
FALSE
1
72001
1
72
001
01804480
Adjuntas
62
-------
Figure 6.1. Map of Puerto Rico plotted from fortified shapefile in R. For this figure, outlying
islands were shifted closer to mainland Puerto Rico.
O
' ~
Box 6.2. R Code for customized mapping of HWBI by county (Fig. 6.2), by domain (Fig. 6.3),
and by indicator (Fig. 6.4).
ft Add HWBI Data to attributes table
hwbi <- read.csv("./Version_PR_US_ByCounty_lndOnly/USind_PR_HWBI.csv", stringsAsFactors=F)
dom <- read.csv("./Version_PR_US_ByCounty_lndOnly/USind_PR_domain.csv", stringsAsFactors=F)
ind <- read.csv("./Version_PR_US_ByCounty_lndOnly/USind_PR_indicator.csv", stringsAsFactors=F)
ft Map HWBI
municipios.HWBI <- merge(municipios.fm, hwbi, by.x = "id", by.y = "FIPS") ft merge attributes with polygons in
fortified data frame
MapHWBI <- ggplot(municipios.HWBI, aes(long, lat, group=group, fill=HWBI)) +
geom_polygon() +
geom_path(colour="gray40", lwd=0.1) + ft adds outline of polygons
coord_equal() +
scale_fill_gradientn(limits=c(min(hwbi$HWBI), max(hwbi$HWBI)),
colours=c("#081d58","#253494","#225ea8","#ld91c0","#41b6c4","#7fcdbb","#c7e9b4","#edf8bl")) +
labs(fill = "HWBI") +
theme(panel.grid.minor=element_blank(),panel.background=element_blank(), ft remove background and gridlines
axis.line=element_blank(), axis.text=element_blank(), axis.line=element_blank(),
axis.title=element_blank(), axis.ticks=element_blank(),
title=element_text(size=14, vjust=2), legend.title=element_blank(), legend.text=element_text(size=8))
MapHWBI
Continued on next page
63
-------
Box 6.2. Continued.
# Map by domain
municipios.DOM <- merge(municipios.fm, dom, by.x = "id", by.y = "FIPS") #polygons x 8 domains
MapDOM <- ggplot(municipios.DOM, aes(long, lat, group=group, fill=DomainScore)) +
geom_polygon() +
geom_path(colour="gray40", lwd=0.1) + # adds outline of polygons
coord_equal() +
facet_wrap(~DOMAIN, ncol=2) + theme(strip.text.x = element_text(size=9)) +
scale_fill_gradientn(limits=c(0.1,0.9),
colours=c("#081d58","#253494","#225ea8","#ld91c0","#41b6c4","#7fcdbb","#c7e9b4","#edf8bl"),
breaks=c(0.1, 0.3, 0.5, 0.7, 0.9), labels=c(0.1, 0.3, 0.5, 0.7, 0.9)) +
labs(fill = "Domain score") +
theme(panel.grid.major=element_blank(), panel.grid.minor=element_blank(),
panel.background=element_rect(colour="black", fill="white"), axis.line=element_line(colour="black"),
strip.background=element_rect(colour="black", fill="white"),
axis.line=element_blank(), axis.text=element_blank(), axis.line=element_blank(),
axis.title=element_blank(), axis.ticks=element_blank(),
title=element_text(size=14, vjust=2), legend.title=element_blank(), legend.text=element_text(size=8),
legend.position="right")
MapDOM
# Map by indicator
municipios.IND <- merge(municipios.fm, ind, by.x = "id", by.y = "FIPS") #polygonsx 25 indicators
municipios.lND$INDICATOR <- as.factor(municipios.lND$INDICATOR)
levels(municipios.lND$INDICATOR) <- sapply(lapply(levels(municipios.lND$INDICATOR), strwrap, width=25), paste,
collapse="\n") # wrap facet wrap labels for long indicator names
MaplNDALL<- ggplot(municipios.lND, aes(long, lat, group=group, fill=IndicatorScore)) +
geom_polygon() +
geom_path(colour="gray40", lwd=0.1) + # adds outline of polygons
coord_equal() +
facet_wrap(~INDICATOR, ncol=3) + theme(strip.text.x = element_text(size=8)) +
scale_fill_gradientn(limits=c(0.1,0.9),
colours=c("#081d58","#253494","#225ea8","#ld91c0","#41b6c4","#7fcdbb","#c7e9b4","#edf8bl"),
breaks=c(0.1, 0.3, 0.5, 0.7, 0.9), labels=c(0.1, 0.3, 0.5, 0.7, 0.9)) +
labs(fill = " Indicator score") +
theme(panel.grid.major=element_blank(), panel.grid.minor=element_blank(),
panel.background=element_rect(colour="black", fill="white"), axis.line=element_line(colour="black"),
strip.background=element_rect(colour="black", fill="white"),
axis.line=element_blank(), axis.text=element_blank(), axis.line=element_blank(),
axis.title=element_blank(), axis.ticks=element_blank(),
title=element_text(size=14, vjust=2), legend.title=element_blank(), legend.text=element_text(size=8),
legend.position-'right")
MaplNDALL
64
-------
>» ' f-k.it• 11 it- fjv\ hr-.Ti
Two versions of the Puerto Rico HWBI mapping are included in this report—within Puerto Rico
(within Puerto Rico version) and between Puerto Rico and the U.S. (U.S. imputations for only
missing indicators version). These versions showcase the differences in the HWBI that can result
from variations in calculation methods resulting from different user needs. Within Puerto Rico
we compared human well-being only among municipios. Metrics were normalized and
standardized relative to the interquartile ranges of Puerto Rico metrics. However, if there was
only one measure for the entire Commonwealth, then data were scaled to the U.S. interquartile
ranges. For the U.S. imputations for only missing indicators version, we compared human well-
being in Puerto Rico to the rest of the U.S. Puerto Rico metrics were normalized and
standardized relative to the interquartile ranges of U.S. metrics and imputed from the U.S. only
if a missing metric resulted in a missing indicator.
HWBI calculated within Puerto Rico (Figures 6.2a) included a greater range of values than the
HWBI based on U.S. imputations for only missing indicators (Figure 6.2b). This is a result of
differences in the normalization and standardization methods previously described. Because
metrics of the within Puerto Rico version were scaled using only the range of Puerto Rico data,
even limited ranges of metric values were scaled 0.1-0.9 and resulted in more variable HWBI,
including some underinflated or overinflated values. Likewise, because the metrics of the U.S.
imputations for only missing indicators version were scaled using a greater range of data, HWBI
was less variable among municipios.
For example, the interquartile range of the metric 'HOMEVAL' (median value of owner occupied
housing units; Appendix A) was narrower for Puerto Rico ($87,000 - $112,475 (A$25,475))
compared to the U.S. ($89,400 - $162,700 (A$73,300)). This is one of two metrics within the
Wealth indicator of the Living Standards domain, and differences between calculation methods
resulted in more variable domain and indicator scores within Puerto Rico than relative to the
U.S. This difference is evident in the comparative maps of the Living Standards domain (Figure
6.3) and Wealth indicator (Figure 6.4). Another reason for differences between versions is that
for some metrics, the interquartile range for Puerto Rico falls outside of the interquartile range
for the U.S. When a metric value is outside of three interquartile ranges, it is considered an
outlier and standardized to the minimum or maximum value, either 0.1 or 0.9. For example, the
interquartile range of the metric 'ALLOFLFE' (percentage of people who experience a
connection to all of life; Appendix A) was 0.95 - 0.97 in Puerto Rico compared to 0.64 - 0.65 in
the U.S. This is the single metric of the Biophilia indicator, which is the single indicator of the
Connection to Nature domain. The differences between calculation methods is evident in the
comparative maps of domain (Figure 6.3) and indicator scores (Figure 6.4), where Connection
to Nature and Biophilia have a range of scores within Puerto Rico but are scaled to the
maximum score of 0.9 when compared to the U.S.
65
-------
Figure 6.2, Maps of mean decadal (2000 - 2010) HWBI for Puerto Rico by municipio where HWBI was calculated a) to compare
HWBI within Puerto Rico (within Puerto Rico) and b) to compare HWBI in Puerto Rico to the rest of the U.S. (U.S. imputations for
only missing indicators). For this figure, outlying islands were shifted closer to mainland Puerto Rico. Higher numbers (light yellow)
indicate higher well-being, and lower numbers reflect lower well-being (dark blue) relative to other municipios.
a)
0.55
b)
66
-------
Figure 6.3. Maps of mean decadal (2000-2010) HWBI domain scores for Puerto Rico by municipio where HWBI was calculated a)
to compare HWBI within Puerto Rico (within Puerto Rico) and b) to compare HWBI in Puerto Rico to the rest of the U.S. (U.S.
imputations for only missing indicators). For this figure, outlying islands were shifted closer to mainland Puerto Rico. Higher
numbers (light yellow) indicate higher well-being, lower numbers reflect lower well-being (dark blue) relative to other municipios.
a)
Connection to Nature
Cultural Fulfillment
Education
Leisure Time
o
Health
Living Standards
Safety and Security
Social Cohesion
0.9
0.7
0.5
0.3
0.1
b/
-------
Figure 6.3. Continued.
b)
Connection to Nature
Cultural Fulfillment
o /-Li':T; "
r
# - -----
Education
Health
r
Leisure Time
Living Standards
s
s
Safety and Security
Social Cohesion
"
<
_
<•
68
-------
Figure 6.4, Maps of mean decadal (2000 - 2010) HWBI indicator scores for Puerto Rico by municipio where HWBI was calculated
a) to compare HWBI within Puerto Rico (within Puerto Rico), and b) to compare HWBI in Puerto Rico to the rest of the U.S. (U.S.
imputations for only missing indicators). For this figure, outlying islands were shifted closer to mainland Puerto Rico. Higher
numbers (light yellow) indicate higher well-being, lower numbers reflect lower well-being (dark blue) relative to other municipios.
Actual Safety
Attitude toward Others
and the Community
Basic Educational
Knowledge and Skills of
Youth
Cultural Activity
Participation
Democratic Engagement
Family Bonding
Basic Necessities
Biophilia
Healthcare
Income
-------
Figure 6.4a. Continued.
Leisure Activity
Participation
Life Expectancy and
Mortality
Lifestyle and Behavior
*
*
Participation and
Attainment
O
fe- f
ft
Perceived Safety
0.9 0.7 0.5 0.3 0.1
Personal Well-being
Physical and Mental
Health Conditions
Social Engagement
Social Support
r
*
Risk
0.9 0.7 0.5 0.3 0.1
70
-------
Figure 6.4a. Continued.
Social, Emotional and
Developmental Aspects
Time Spent
Wealth
4ML *
* jjjSfpl 4j
/
Work
Working Age Adults
71
-------
Figure 6.4b. Continued.
b)
Actual Safety
c?
Basic Necessities
Cultural Activity
Participation
Attitude toward Others
and the Community
Biophilia
o
Democratic Engagement
Basic Educational
Knowledge and Skills of
Youth
o
0.9 0.7 0.5 0.3 0.1
Family Bonding
Healthcare
o r
r
Income
0.9 0.7 0.5 0.3 0.1
72
-------
Figure 6.4b. Continued.
Leisure Activity
Participation
¦ -S*-
Participation and
Attainment
Life Expectancy and
Mortality
Perceived Safety
Lifestyle and Behavior
0.9 0.7 0.5 0.3 0.1
Personal Well-being
o
Physical and Mental
Health Conditions
Risk
• WM
*
Social Engagement
Social Support
o
0.9 0.7 0.5 0.3 0.1
73
-------
Figure 6.4b. Continued.
Social, Emotional and
Developmental Aspects
Time Spent
Wealth
Work
Working Age Adults
0.9 0.7 0.5 0.3 0 1
r
¦¦¦¦
74
-------
Chapter 7: Comparing HWBI Across Time
poiral Flexibility of HWBI
The HWBI can be calculated for a single time period (e.g., mean decadal) or across time (e.g.,
annual time series). Temporal flexibility was incorporated into the HWBI in order to evaluate
well-being overtime in relation to changes in social, economic, and environmental services. As
a result, the index has the potential to measure the sustainability of alternative policies and
decision alternatives that affect human well-being (Smith et al., 2014a; Summers et al., 2014).
7.2. Evaluation of HWBI Time Series
Overall, the HWBI for Puerto Rico increased during the annual time series of 2000-2013 (Figure
7.1). Although all 78 municipio HWBIs increased during the time series, some changed little
(e.g., Vieques, Culebra, Canovanas, Catano) while others changed considerably (e.g., Las Marias,
Maricao, Jayuya, Juana Diaz) (Figures 7.2 and 7.3). Many municipios showed a decline in the
mid-2000's that may correlate with the national economic recession that contracted the
economy of Puerto Rico (Bram et al., 2008). The increasing trend in the HWBI over time is
largely consistent with the HWBI indictor time series, however some indicators consistently
declined (e.g., Physical and Mental Health Conditions, Working Age Adults) (Figure 7.4).
In practice, the ability to evaluate the HWBI over time depends on the temporal availability of
data. For Puerto Rico, 20 metrics were based on data for only a single year. This is evident in
the time series of HWBI indicators based on a single metric value, as they remain constant over
time (Figure 7.4). For example, Biophillia remains constant because it was measured using only
a single metric (ALLOFLFE) for a single year (2001). Conversely, Actual Safety is based on
multiple metrics with a range of temporal availability (Appendix A), increasing the variability
and annual fluctuations of this indicator.
75
-------
Figure 7.1. Time series for Puerto Rico HWBI (U.S. imputations for only missing indicators).
Higher numbers indicate higher well-being, and lower numbers reflect lower well-being.
0.60
Or- CNCOTIOCOI-.COCDOT— CM CO
OOOOOOOOOOt-^-t-t-
oooooooooooooo
CMCMCNCMCMCMCslCNJCMCMCNCMCMCN
Year
Figure 7.2. Net change in Puerto Rico HWBI by municipio for 2000 - 2013 (U.S. imputations
for only missing indicators). For this figure, outlying islands have been shifted closer to
mainland PR. Higher numbers (dark orange) indicate larger increases in HWBI score over time;
numbers closer to zero (light yellow) indicate only slight increases over time. No municipio
showed a net decline in well-being, as would be indicated by a negative change.
76
-------
Figure 7.3. Time series for Puerto Rico HWBI by municipio (U.S. imputations for only missing indicators).
X
05
T3
_C
U)
c
"05
_Q
05
5
C
03
E
X
0.60 '
0,55 *
0.50 -
0.45 ¦
0 60 ¦
0 55 •
0.50 •
0 45 •
0.60
Adjuntas
Aguada
Aguadilla
Aguas Buertas
Aibonito
Ariasco
Arecibo
Arroyo
Barcelorteta
Barranquitas
•
• » »"* *
Bayamon
Cabo Rojo
Caguas
Camuy
Canovanas
Q 30 •
0.55 ¦
0 50 5
0.45 ¦
Carolina
Catano
Cayey
Ceiba
Ciates
1.60 ¦
) 55 5
3 50 ¦
3 45 ¦
,150 5
J.55 •
J.50 5
) 45 .
J 60 ¦
55 ¦
.50 ¦
.45 ¦
i 60 •
55 ¦
50 ¦
.45 ¦
Cidra
Coamo
Comerio
Corozal
Culebra
x
~*¦»» » 1
Guayanilla
Guaynabo
Gurabo
Hatillo
Hormigueros
Humacao
Isabela
Jayuya
Juara Diaz
Junccs
cj oj c*-« cm
Year
77
-------
Figure 7.3. Continued.
0.60"
0 55"
0.50-
045*
e.oo ¦
0,56 ¦
0.50 ¦
0.45 '
Lajas
Lares
Las Marias
Las Piedras
Loiza
Luquillo
Manati
Maricao
Maunabo
Mayaguez
-
Moca
Morovis
Naguabo
Naranjito
Orocovis
x
CD
xy
_c
CD
C
'(D
X!
-J- 0.00 !
> 0,55 !
^ 0 50 •
2 0.45 ¦
oeo •
0 55 ¦
0.50 •
0.45 ¦
0.60 5
0 55 !
0.50 •
0.45 •
0 60 5
0 55 •
0 50 •
0 45 •
Patillas
Periuelas
Ponce
Guebradillas
Rincon
Rio Grande
Sahara Grande
Salinas
San German
San Juan
»~*•« • ¦ *•>
x
©
T3
en
c
£s
San Lorenzo
»~*»»! * *¦'
San Sebastian
Santa Isabel
Toa Alta
Toa Baja
_ 0,00 •
5 0,55 •
0,50 •
0 45 •
Trujillo Alto
Utuado
Vega Alta
Vega Baja
Vieques
c
CO
E
3
X
0,00
0,55
0,50
0 45
Villalba
Yabucoa
Yauco
Year
cm eg
Year
o
o
g
f-4
"S3-
S
CJ
CO
g
CM
0
1
C^4
oo
g g
CJ CJ
Year
C
s
T
i
Cv
CO
g
CJ
78
-------
Figure 7.4. Time series for Puerto Rico HWBI indicator scores (U.S. imputations for only missing indicators).
Actual Safety
Cultural Activity
Participation
Leisure Activity
Participation
Attitude toward
Others and the
Community
Democratic
Engagement
Life Expectancy and
Mortality
Basic Educational
Knowledge and
Skills of Youth
Family Bonding
/
Lifestyle and
Behavior
^*<1 «»»»»>~
Basic Necessities
Biophilia
Healthcare
Participation and
Attainment
«>«»»»
Income
Perceived Safety
O T
8 S
CM CM
CO
8
o o
C-4 CM
8
c-g
CO
8
o
CM
O
c-g
o
CM
o
CM
Year
79
-------
Figure 7.4. Continued.
Personal Well-being
Physical and Mental
Health Conditions
Social, Emotional
and Developmental
Aspects
Time Spent
„ /*—
~YS
BP'WWW V w WWW WWW WW
«
o
CM
«
o
CNi
Risk
••••••
Wealth
Social Engagement
Work
Social Support
A. MkJtL^kJtk
Working Age Adults
W^w
Tt* CO
8 8
CM CM
Year
CO
O
CM
CO
o
CM
CO
o
C-4
80
-------
Chapter 8: Comparing HWBI Among Populations
8.1. Considerations for HWBI Comparison Among Populations
Various adaptations, including metric selection, imputation of missing data, and
standardization, must be considered when developing a HWBI that allows comparisons
between populations. Data can be challenging to obtain for Puerto Rico (Bradley et al., 2016),
and federal agencies and national surveys often focus efforts on the 50 states. To determine
how to preserve the distinct well-being "fingerprint" of Puerto Rico while maintaining a
comparative index, we evaluated alternative treatments for missing metric data including
metric substitution, removal, and imputation. These alternatives are supported by a previous
HWBI adaptation that compared Native American populations to the national index while
retaining unique distinctions (Smith et al., 2014b).
I vhMii-Ti l vlternative Methods for Missing Metrics
Thirty-one metrics were not available for Puerto Rico and were either substituted, imputed, or
removed for HWBI calculation. We examined the difference between substitution and
imputation for 15 metrics where we identified a different but similar measure that captured the
scope of the U.S. metric and had comparable units for normalization and standardization. We
found that individual metrics differed in magnitude when substitutions were compared to
imputations, but there was no consistency to which method resulted in higher or lower metric
values (Figure 8.1).
Four additional substitute metrics were not suitable to a U.S. comparison (Figure 8.2). Metrics
based on substituted measures were not suitable to a U.S. comparison because they were
based on an alternative geography (LEISURE) or had different units that could not be
standardized to the U.S. (PERARTS, SOVI, and VACATION). Substituted data for SOVI and
VACATION were assigned a median value of 0.5 because these meters were based on a single
observation that could not be standardized within Puerto Rico or to the U.S.
81
-------
Figure 8.1. Comparison of mean decadal (2000-2010) metric values for Puerto Rico calculated
using substituted measures and U.S. imputations.
0.9 -
0.7 -
(D
3
(U
>
o
© 0 5 -
£
£
0 3 -
Substituted |Q| Imputed from U.S.
Metric
Figure 8.2. Comparison of substituted and imputed metrics for Puerto Rico.
0.9 "
0 1 -
Substituted Qj Imputed from U.S.
LEISURE PERARTS SOVI VACATION
Metric
82
-------
We also explored the alternative of removing missing metrics from the Puerto Rico HWBI to
avoid any contributions that were not specific to Puerto Rico. However, in order to maintain
the integrity of the index (Smith et al., 2014b), we did not remove metrics that resulted in the
elimination of an indicator and imputation from the U.S. was necessary. Overall, metric removal
resulted in a lower mean decadal (2000-2010) HWBI (0.46) than substitution (0.51) (Figure 8.3).
Metrics based on data specific to Puerto Rico were expected to better reflect the population
due to the distinct historical, cultural, social, environmental, and economic influences on
human well-being. Therefore, when comparing the HWBIs, we used Puerto-Rico specific data as
much as possible to maintain its local distinction and only imputed data from the U.S. if removal
resulted in the elimination of an indicator.
0.55
X
a
CD
E
Substituted Q| Removed
Figure 8.3. Comparison of mean decadal (2000-2010) HWBI for Puerto Rico calculated with
substituted (72 metrics) and without substituted measures (60 metrics with U.S. imputations
only for missing indicators).
It was necessary to impute missing metrics when a suitable substitute was not available in
order to compare HWBIs between Puerto Rico and the U.S. We followed the U.S. HWBI method
of stratification of county-level data based on RUCC-GINI combination to impute missing
metrics, which accounts for similarities in well-being due to spatial relationships with large
urban centers and the dispersion of income (Smith et al., 2014a). Alternative methods were
developed based on the number of imputations (all missing metrics or only when removal
resulted in the elimination of an indicator) and source of imputations (Hawai'i RUCC or U.S.
RUCC-GINI) (Table 4.1). The U.S. state of Hawai'i was considered a plausible imputation source
because it is also geographically separated and culturally distinct from the contiguous states
and faces similar economic challenges and opportunities as an island chain.
83
-------
To evaluate these imputations, we compared interquartile ranges of Puerto Rico, Hawai'i, and
the U.S. (Figure 8.4). For most metrics, Puerto Rico interquartile ranges were outside of the U.S.
interquartile ranges (e.g., CANCER), although some were similar to the U.S. (e.g., MATHTEST).
Imputations based on all U.S. RUCC-GINI combinations were conceptually preferable to Hawai'i
because only a few metrics were more similar between Puerto Rico and Hawai'i than Puerto
Rico and the U.S (e.g., ADULTLIT). Furthermore, there is a lack of RUCC-GINI combinations
because Hawai'i only has five counties compared to 78 Puerto Rico municipios.
Analysis of Variance (ANOVA) was used to test whether calculated HWBI differed significantly
by imputation method. Tukey's pairwise post-hoc comparisons were used to investigate which
imputation methods differed from each other. Puerto Rico mean decadal (2000-2010) HWBI by
municipio differed among imputation methods (ANOVA, p<0.001) (Figure 8.5) and Tukey's
pairwise post-hoc comparisons revealed differences among versions. When all missing metrics
were imputed, the U.S. imputation produced significantly higher estimates of HWBI than
Hawai'i imputations. However, when only metrics that would result in a missing indicator were
imputed, there was no difference between the U.S. and Hawai'i imputations (Figure 8.5). For
the domain scores, only Leisure Time (greater with U.S. imputations for all missing metrics) and
Safety and Security (greater with Hawai'i imputations for all missing metrics) differed among
these two versions (Figure 8.6).
Despite basing imputations on geographic and sociodemographic similarities (degree of
urbanization and income inequality), there are distinct historical, cultural, social,
environmental, and economic differences among populations that may have a greater influence
on human well-being than demographic similarity. As such, we prioritized the inclusion of
Puerto Rico-specific data through metric substitution whenever available and imputed data
from the U.S. only when removal resulted in the elimination of an indicator (Figure 4.2). This
resulted in the reduction of metrics (80 metrics used to calculate U.S. HWBI compared to 72
metrics to calculate Puerto Rico HWBI) without eliminating indicators or domains. These 72
Puerto Rico metrics were normalized using U.S. interquartile ranges to identify outliers and
standardized from 0.1-0.9 using the same scaling methods as the U.S.
84
-------
Figure 8.4. Comparison of unstandardized metric values for Puerto Rico, Hawai'i, and the United States.
so -
60 ~
ACCMM
ADLTASTHMA
0.14 -
0,12 -
ADULTLIT
I"
0,15 -
5
I
7 -
0.10 ~
0.10 -
0.08 -
ALCOHOL
0.12 -
0.10 -
0.08 -
Q.06 -
0.04 -
"T"
T 1
! 1 f
OS CO f-
o o o
ALLOFLFE
CD
O)
c
CD
0>
'¦E
CD
cr
B
c
0.50 ¦
0.25 ¦
0.00 •
-0.25 ¦
-0.50 ¦
0.5 ¦
0.4 '
0.3 ¦
0.95 '
0 90 -
0.85 '
0.80 ¦
0.75 ¦
ASTHMORT
CANTRUST
CHLDSOCIAL
0.8 •
0.7 •
0.8 ¦
0.95 •
0 90 ¦
0.80 1
0.75 ¦
BEAUSPRT
CHLDACTV
CITYSATiS
BULLY
3Z x
CHLDASTHMA
CLSETOWN
1.0 •
0.8 '
o.e ¦
0.4 1
CANCER
CHLDHLTH
CLSFRNDFAM
0.24 •
0.22 '
0.20 ¦
0.18 ¦
2.4 •
2.2 ¦
2.0 ¦
1.8 ¦
0.15 ¦
0.14 •
0.13 '
0.12 ¦
0.11 '
CANCMORT
CHLDREAD
!
CONFACT
CD
CO
>
G
0.20 •
0.16 «
0.12 •
DEPRESSION
0.20 ¦
0.15 ¦
0.10 '
DIABETES
DIABMORT
0.100 ¦
0,075 ¦
0.050 ¦
DISCRM2
0.82
0.80 '
0.78 1
0.78
0.74
EMTSUPRT
0)
C
CO
~o
0
0.925 •
0.900 '
0.875 •
0.850 •
0.825 '
0.800 •
0.50 ¦
0.45 •
0.40 -
0.85 *
FAMDOC
HELPFUL
0.85
0.80
0.75 ¦
FOODSECURE
HOMEAFFORD
0,64 -
0.60 -
0.5© -
0.52 -
GRPACTV
HOMEVAL
0.08 ¦
0.06 ¦
0.04
0.02 ¦
HAPPY
HRT ATTACK
60
50 ¦
40
0.100
0.075
0.050
HBI
HRTDISEASE
0.28 •
0.26 ¦
0.24 •
HRTDISMORT
0
-
u
KH
»»
lcs
M
m -
0
eb
HSGRAD
INFMORT
Puerto Rico Hawaii
0,975 ¦
0.950 '
0.925 ¦
0.900 ¦
Q.875 '
JOBLOSE
United States
0.9
0,8
0.7 ¦
0.6
0.5 ¦
JOBSATIS
85
-------
Figure 8.4. Continued.
312 ¦
308 '
304 ¦
300 '
LEISURE
LIFESATiS
$
LIFEXPCT
0.350 ¦
0.325 '
0.300 ¦
0.275
©
O)
c
cn
m
0.27 -
0.24 H
•c.
0.21 -
CO
z>
cr
©
n" -
r
0 4 J
6.3 H
r
0.2 ™
£
©
0.85
CO
0,80 -
>
0,75 -
Q
0.70 -
i_
CD
E
9 -
c
8 -
7 -
CO
¦o
©
2
0.09 '
0.06 ¦
0.03 •
MEALS
OBESITY
POVERTY
READTEST
SENiORCARE
TEENSMK
50000 ¦
40000 '
30000 •
20000 ¦
1.00 •
0.75 '
0.50 ¦
0.25 •
0.00 •
1.0 ¦
0.9 ¦
0.8 1
0.7 •
MEDiNCOME
PARTNEDU
POVPERSIST
REGVOTRS
SOVi
TOTRATE
1.0 •
0.8 '
0.8 •
0.6 -
0.8
Q.7
0.6
MTGDEBT
PERARTS
PRCVDHLTH
SATDEM
STROKE
TRUSTGOV
300 ¦
200 •
VIOLCRIME
"T"
.Jit
0.50 ¦
0.45 ¦
0.40 '
0.35 ¦
0.30
0.2b 1
VOICENGOV
0.50
0.45
0.40
VOLNTR
Puerto Rico * Hawaii
86
0.16 •
0.12 •
0.08 '
0.04 ¦
LONGWRKHRS
0.775 '
0.750 ¦
0.725 ¦
0.700 •
MATHTEST
NATHAZHLOSS
NORMWRKHRS
0.8 ¦
0.7 ¦
0.6 •
0.5 ¦
PHYSACTIV
0.8 •
0.7 '
0.6 ¦
0.5 •
0.4 '
POLINTRST
0.70 ¦
0.65 '
0.80 ¦
0.55 •
PRCVDSAFE
SATiSHLTHC
4000 ¦
3000 '
2000 ¦
1000 ¦
0.80
0.75
0.70 ¦
PROPCRIME
SC1TEST
SUICDMORT
j
TEENPREG
UNIVGRAD
4.7
4.5 '
4,3
VACATION
0.8 ¦
0.7 '
0.6 '
0.5 ¦
VOTRTOUT
0.50
0.45
0.40
0.35
0.30
WATCHTV
United States
-------
Figure 8.5. Comparison of mean decadal (2000-2010) HWBI for Puerto Rico using alternative
methods based on number of imputations (all missing metrics or only when a missing metric
resulted in a missing indicator) and source of imputations (Hawai'i by RUCC or U.S. by RUCC-
GINI). Error bars give standard error across municipios.
0.55
0.50
0.45
0.40
U.S.
imputations
for all missing
metrics
Hawaii
imputations
for all missing
metrics
Version
U.S.
imputations for
only missing
indicators
Hawaii
imputations for
only missing
indicators
87
-------
Figure 8.6. Comparison of mean decadal (2000-2010) domain scores for Puerto Rico using
alternative methods based on number of imputations (all missing metrics or only when a
missing metric resulted in a missing indicator) and source of imputations (Hawai'i by RUCC or
U.S. by RUCC-GINI). Error bars give standard error across municipios.
Connection
to Nature
Cultural Education
Fulfillment
Health Leisure
Time
Domain
Living Safety and Social
Standards Security Cohesion
U.S. imputations for all missing metrics
Hawaii imputations for all missing metrics
j U.S. imputations for only missing indicators
Hawaii imputations for only missing indicators
88
-------
8.3. Comparison of Puerto Rico and U.S. HWBI
Puerto Rico ranks in the lowest quartile (0.22 percentile) with a mean decadal (2000-2010)
HWBI of 0.51 compared to 0.50-0.56 for U.S. states (Figure 8.7). Municipios range from 0.27-
0.61 percentile ranking compared to U.S. county equivalents when evaluating Puerto Rico and
the U.S. at the county level (Figure 8.8). Over time, Puerto Rico HWBI follows the U.S. with a
slight upward trend (Figure 8.9).
X 0.4 ¦
0
"O
0
_Q
03
E
i i i i i i i i i i i i i i i i i i i i i i i i i i i i i r
i i i i i i i i i i r
i i i i i i i i r
ihzt:liJirliJ<^di-Q:zoi><oz>o<
Z3^>WZ12->>Zl
-------
Figure 8.8. Mean decadal (2000-2010) HWBI percentile rankings for 78 municipios of Puerto Rico (U.S. imputations for only
missing indicators) compared to U.S. county equivalents (n=3,221).
0.6
>,
c
GO
3
O
3
o
O
>>
XI
X
»
0)
TJ
CO
c
a
E
CO
o
c
o
0)
•W
.a
a>
t
'£5
a
®
o
u
c
a>
CO
cl E
3
X
0.2 L
Municipio
90
-------
Figure 8.9. Annual time series (2000-2010) of HWBI values for U.S. states, Washington, D.C.,
and Puerto Rico (U.S. imputations for only missing indicators).
AK
At
AR
AZ
CA
[ > ¦
«»|t»|*
CO
CT
DC
FL
>-« » »*•» »-* * * *
»••~••<
GA
IA
KY
LA
MM'
MD
X
©
"O
_c
CO
c
"©
.0
C
CO
E
3
X
Ml
MS
MT
NC
ND
NE
******
i»»t * * *
NH
NJ
NV
NY
!>>« »•«
*»>*-¦
OH
OK
OR
PA
PR
• »
sc
SD
TN
TX
UT
VA
VT
WA
Wl
wv
WY
C\ *T 'X'
« M M CV CM fsj C4 (Ni Csl CM fNi CV CM CM CM CM CN« CM
Year
Year
Year
t i r
«w/ ' W ^ ^
55 CO O
Year
Year
91
-------
Among domain and indicator scores, Puerto Rico has a unique "fingerprint" of well-being assets
and deficits relative to the U.S. To facilitate comparison between the 8 domain and 25 indicator
scores of Puerto Rico and the U.S. state equivalents, Puerto Rico was compared to the U.S.
states with highest (New Hampshire) and lowest (Louisiana) mean decadal (2000-2010) HWBI
(Figure 8.10 and Figure 8.11). Puerto Rico had considerably higher Connection to Nature
(comprised of the Biophilia indicator) and moderately higher Cultural Fulfillment (comprised of
the Cultural Activity Participation indicator) compared to both New Hampshire and Louisiana.
This is in contrast with the Living Standards domain, which was considerably lower in Puerto
Rico than both U.S. states. This trend was paralleled by the indicator scores within this domain
(Basic Necessities, Income, and Wealth), although Work was similar among areas. Disparity
among living standards is reflected in the continued migration of people from Puerto Rico to
the U.S. mainland. Along with the economic gap between Puerto Rico and the mainland U.S.,
Puerto Rico has seen an increased loss of jobs and disinvestment in industry since the Great
Recession (Abel and Deitz, 2014; Bram et al., 2008).
Puerto Rico
New Hampshire
Louisiana
(C
E
o
Q
0.9 -
0.7 -
0.5 -
0.3 -
0.1 -
Connection to Nature
Education
Health
Cultural Fulfillment
Leisure Time
Living Standards
Social Cohesion
Safety and Security
Figure 8.10. Comparison of mean decadal (2000-2010) domain scores for Puerto Rico (U.S.
imputations for only missing indicators) and the U.S. states with highest (New Hampshire)
and lowest (Louisiana) HWBIs.
92
-------
Figure 8.11. Comparison of mean decadal (2000-2010) indicator scores for Puerto Rico (U.S.
imputations for only missing indicators) and the U.S. states with highest (New Hampshire)
and lowest (Louisiana) HWBIs.
Puerto Rico
New Hampshire
Louisiana
o
o
o
"co
o
~o
0.3 ¦
0.1 -
1
I
Biophilia
Basic Educational Knowledge and Skills of Youth
Participation and Attainment
Social, Emotional and Developmental Aspects
Healthcare
Life Expectancy and Mortality
Lifestyle and Behavior
Personal Well-being
Physical and Mental Health Conditions
Cultural Activity Participation
Leisure Activity Participation
Time Spent
Working Age Adults
Basic Necessities
Income
Wealth
Work
Attitude toward Others and the Community
Democratic Engagement
Family Bonding
Social Engagement
Social Support
Actual Safety
Perceived Safety
Risk
93
-------
For the Health domain and its indicators, especially Healthcare and Life Expectancy and
Mortality, Puerto Rico was similar to Louisiana but moderately lower than New Hampshire. This
was also the case for the Education, Social Cohesion, and Safety and Security domains, although
for these domains individual indicator trends varied. In Education, Puerto Rico had considerably
lower Social, Emotional, and Developmental Aspects but considerably higher Basic Educational
Knowledge and Skills of Youth than both states. Puerto Rico also had moderately higher
Participation and Attainment than Louisiana. For Social Cohesion, Puerto Rico had considerably
higher Social Support and slightly higher Democratic Engagement but considerably lower Family
Bonding than both states. Also, Puerto Rico had moderately lower Attitude toward Others and
the Community and Social Engagement compared to New Hampshire. Similarly for Safety and
Security, Perceived Safety for Puerto Rico was considerably lower than both states, while Actual
Safety was considerably higher than Louisiana. Risk was similar among all areas. The Leisure
Time domain score for Puerto Rico was similar to both New Hampshire and Louisiana. However,
among indicators Puerto Rico had considerably lower Leisure Activity Participation, moderately
higher Working Age Adults, and similar Time Spent as both states.
The apparent contradictions among indicators in a domain reflect the complexity of well-being.
This is attributed to a diversity of components used that include both subjective and objective
measures to calculate HWBI. For example, Puerto Rico had a lower perception of safety but
similar or even higher actual safety compared to U.S. states. This may be the result of highly
publicized murder rates in San Juan, largely due to drug trafficking, which may have incited
distrust in public safety (Caribbean Business, 2014a).
The comparison between Puerto Rico and U.S. states may also reveal limitations in the Puerto
Rico data. For example, in Puerto Rico HWBI Family Bonding was based solely on television
watching by children whereas the U.S. indicator was based on television watching, time spent
reading to children, and meals eaten as a family (Appendix A). These metrics may not actually
account for the value of extended family in Hispanic culture (Tienda and Mitchell, 2006). Other
unique attributes of Puerto Rican culture, such as traditional cooking or informal neighborhood
gatherings, and the Puerto Rican underground economy, including street vendors, roadside
stands, and cash only businesses, may also not be reflected in national-scale statistics.
94
-------
8.4. Comparison of HWBi among San Juan Bay Estuary Watershed
Municipios
The San Juan Bay Estuary is part of the EPA's National Estuary Program and is designated as "an
estuary of national significance" (Villanueva, 2000). The watershed contains the densely
populated greater San Juan metropolitan area and includes area from eight municipios: Toa
Baja, Catano, Bayamon, San Juan, Guaynabo, Carolina, Lofza, and Trujillo Alto (Figure 8.12).
Industrial pollution associated with coastal population growth and development threatens the
estuary itself along with economic and recreational services dependent on this natural resource
(Villanueva, 2000).
a)
„ s , _ .J
i \ j ) X, V
San Juan Bay Estuary Watershed
* w-
I
j
K > / ^
HLoiza0Catano0Toa Baja0Carolina0San JuanHTrujillo AltoMBayam°nHGuaynabo
Figure 8.12. Map of a) Puerto Rico's San Juan Bay Estuary Watershed and b) eight associated
municipios. HWBI was calculated at the municipio level, including area outside of the
watershed boundary. For this figure, outlying islands have been shifted closer to mainland PR.
95
-------
The eight municipios associated with the San Juan Bay Estuary Watershed included both the
lowest (Lofza and Catano), moderately low (Toa Baja), and highest (Carolina, San Juan, Trujillo
Alto, Bayamon, and Guaynabo) HWBI compared to municipios across Puerto Rico (Figure 8.13).
The disparity among these highest and lowest ranked municipios of the San Juan Bay Estuary
Watershed was driven largely by differences in Education, especially Basic Educational
Knowledge and Skills of Youth and Participation and Attainment indicators (Figure 8.14).
Additionally there were differences in Living Standards (Income and Wealth), Health (Physical
and Mental Health Conditions), and Leisure Time (Leisure Activity Participation and Time
Spent). Conversely, Social Cohesion (Social Engagement) was greatest in the lowest ranked
municipio of Lofza. This comparison, although limited by the availability of municipio-specific
data (Appendix A), reveals differences in human well-being within the greater San Juan area
that could be investigated further with an HWBI based on even finer spatial resolution.
oj •
0.4 -
® 0.2 ¦
cc o
NSC
at 33 in 's> ® 's> ^ c
JSr9i ro R
,15 spill
< a 1r,OS
I *
U
C C
(j'-JCt J
-------
Figure 8.14. Puerto Rico mean decadal (2000-2010) HWBI indicator scores for eight municipios of San Juan Bay Estuary
Watershed (U.S. imputations for only missing indicators).
Actual Safety
Attitude toward
Others and the
Community
Basic Educational
Knowledge and
Skills of Youth
Basic Necessities
Biophilia
4
Ik
Cultural Activity
Participation
Democratic
Engagement
Family Bonding
Healthcare
Income
*
jLoiza Catano Toa Baja Carolina San Juan Trujillo AltoBBayamonHjGuaynabo
97
-------
Figure 8.14. Continued.
CD
O
O
U)
0.9-
0.7-
0.5-
0.3-
0.1 ¦
03
O
TD
C
0.9-
0.7-
0.5-
0.3-
0.1 -
0
O
O
(/)
_o
TD
c
0.9 ¦
0.7-
0.5-
0.3
0.1 ¦
Leisure Activity
Participation
Personal Well-being
Social, Emotional
and Developmental
Aspects
Life Expectancy and
Mortality
Physical and Mental
Health Conditions
Time Spent
Lifestyle and
Behavior
Wealth
Participation and
Attainment
Social Engagement
Work
Perceived Safety
Social Support
Working Age Adults
Lofza Catano Toa Baja Carolina San Juan Trujillo AltoBjBayamonBGuaynabo
98
-------
Chapter 9: Summary
HWBI integrates measures of economic, social, and environmental contributions to well-being.
Responsive to the flows of goods and services within the natural and built environments, HWBI
can serve as an endpoint of human well-being and indicate sustainability when tracked over
time. Therefore HWBI can facilitate the evaluation of alternative resource management and
policy decisions within formal decision making frameworks (Smith et al., 2014b; Summers et al.,
2014). Community decisions (e.g., funding decisions, school sitings, emergency preparedness
activities, public works projects) inevitably involve choices and tradeoffs among the availability
and quality of economic, social, and environmental goods and services that impact the well-
being of society. HWBI can provide a way to quantify the well-being objectives of constituents,
so they can be more fully considered in decisions. A school siting in a rural area to improve
access to educational services, for example, may lead to increases in education but also social
cohesion, safety and security, health, and other components of well-being (Summers et al.
2016). Tracked over time, HWBI can be used to evaluate the success of alternative policy
decisions and inform future decisions (Smith et al., 2014b; Summers et al., 2014).
We adapted the HWBI to compare human well-being both within Puerto Rico and to the U.S.
population. This adaptation to Puerto Rico, a territory of the U.S. with commonwealth status,
demonstrates the transferability of HWBI to institutionally similar but culturally and
geographically distinct population of the U.S. Previous adaptations of the index have
demonstrated the scalability of the index to a specific population (American Indian Alaska
Native population of the U.S.; Smith et al., 2014b)). Because HWBI was developed from
domains and indicators considered applicable to all populations, the Puerto Rico HWBI may also
support the applicability of HWBI outside of the U.S.
Puerto Rico HWBI builds on the existing HWBI framework and allows for Puerto Rico to be
included in local- and national-level decision and policy making applications of the index.
Although data availability varied, we maintained the integrity of the index for comparisons with
the U.S and measured well-being distinctive to Puerto Rico. For this adaptation, we scaled
HWBI to the 78 municipios (municipalities) which make up Puerto Rico and function as county-
equivalents. The ability to scale HWBI to a community or population using either data collected
for a specific population or imputed from broader (e.g., national, state, regional) to finer spatial
scales (e.g., county, neighborhood, socioeconomic group) offers an opportunity to further
understand well-being within Puerto Rico. In addition to scalability, HWBI can be customized
using community-specific measures of well-being to better reflect a specific population.
EPA's Sustainable and Healthy Communities Research Program (EPA 2015) is developing
research, data, and tools to expand the capabilities of communities to consider the social,
economic, and environmental impacts of decision alternatives on community well-being.
Research includes efforts to link delivery of ecosystem goods and services to human well-being.
Indices, such as HWBI, can help communities to assess and track well-being, and identify
decisions and interventions that can contribute to and promote sustainable well-being.
99
-------
References
Abel, J.R., and R. Deitz. 2014. The Causes and Consequences of Puerto Rico's Declining
Population. New York: Federal Reserve Bank of New York.
Bivand, R., T. Keitt, B. Rowlingson, E. Pebesma, M. Sumner, R. Hijmans, and E. Rouault. 2015.
Package 'rgdal' version 1.0.6: Bindings for the Geospatial Data Abstraction Library. R
Project, CRAN Repository, http://cran.r-proiect.org/web/packaees/redal/redal.pdf
Bradley, P., W. Fisher, B. Dyson, S. Yee, J. Carriger, G. Gambirazzio, J. Bousquin, and E. Huertas.
2016. Application of a Structured Decision Process for Informing Watershed
Management Options in Guanica Bay, Puerto Rico. U.S. Environmental Protection
Agency, Office of Research and Development, Narragansett, Rl. EPA/600/R-15/248.
Bram, J., F.E. Martinez, and C. Steindel. 2008. Trends and Developments in the Economy of
Puerto Rico. Current Issues in Economics and Finance 14:1-7.
Budlender, D. 2008. The Statistical Evidence on Care and Non-Care Work across Six Countries.
Geneva: United Nations Research Institute for Social Development, Gender and
Development Programme.
Caribbean Business. 2011. Nearly one-third of Puerto Ricans know all their neighbors.
September 29, 2011.
Caribbean Business. 2013. Survey: PR boasts one of highest levels of job commitment in world.
October 20, 2013.
Caribbean Business. 2014a. Most Puerto Ricans perceive crime has gone up. February 27, 2014.
Caribbean Business. 2014b. Percentage considering moving out of Puerto Rico continues to
increase. August 28, 2014.
Centers for Disease Control and Prevention (CDC). 2011. Asthma in Puerto Rico. Fact Sheet,
National Center for Environmental Health, Division of Environmental Hazards and Health
Effects.
Centers for Disease Control and Prevention (CDC). 2015a. 1991 - 2013 High School Youth Risk
Behavior Survey Data. Accessed January 15, 2015 from
http://nccd.cdc.eov/vouthonline/
Centers for Disease Control and Prevention (CDC). 2015b. Behavioral Risk Factor Surveillance
System (BRFSS) Data (ASCII). Accessed December 8, 2014 from
http://www.cdc.gov/brfss/annual_data/annual_data.htm
Centers for Disease Control and Prevention (CDC). 2015c. Invasive Cancer Incidence—Puerto
Rico, 2007-2011. Atlanta: Morbidity and Mortality Weekly Report (MMWR) 64:389-
393.
100
-------
Centers for Disease Control and Prevention (CDC). 2015d. National Center for Health Statistics
(NCHS), National Vital Statistics System, Birth Data Files. Accessed March 17, 2015 from
http://www.cdc.gov/nchs/data access/Vitalstatsonline.htm
Centers for Disease Control and Prevention (CDC). 2015e. National Center for Health Statistics
(NCHS), National Vital Statistics System, Multiple Cause of Death Data. Downloaded
from the National Bureau of Economic Research. Accessed July 14, 2015 from
http://www.nber.org/data/vital-statistics-mortalitv-data-multiple-cause-of-death.html
Central Intelligence Agency (CIA). 2015. CIA World Factbook.
http://www.cia.gov/library/publications/the-world-factbook/geos/rq.html
Comision Estatal de Elecciones. 2015. Election Turnout by Senate Districts and Precincts.
Downloaded from Alvarez-Rivera, M. Elecciones en Puerto Rico. Accessed February 4,
2015 from http://electionspuertorico.org/
Compama de Turismo de Puerto Rico (CTPR). 2015. Estudios del Mercado (Tourism Company of
Puerto Rico [PRTC], Market Studies). Downloaded from Instituto de Estadfsticas de
Puerto Rico. Accessed June 26, 2015 from
http://www.estadisticas.gobierno.pr/iepr/Estadisticas/lnventariodeEstadisticas/tabid/l
86/ctl/view_detail/mid/775/report_id/123d2077-0571-4582-a4bf-
cf5468e49dfd/Default.aspx
Environmental Protection Agency (EPA). 2015. Sustainable and Healthy Communities: Strategic
Research Action Plan 2016-2019. EPA Report EPA601/K-15/006.
Federal Bureau of Investigation (FBI). 2015. Uniform Crime Reports. Accessed November 19,
2014 from httpV/www.fbi.gov/about-us/ciis/ucr/crime-in-the-u.s
Gall, M. 2007. Indices of Social Vulnerability to Natural Hazards: A Comparative Evaluation.
University of South Carolina: Department of Geography.
Gasparini, L., W.S. Escudero, M. Marchionni, and S. Olivieri. 2010. Multidimensional poverty in
Latin America and the Caribbean: New evidence from the Gallup World Poll.
Universidad Nacional de La Plata: Center for Distributional, Labor and Social Studies
(CEDLAS).
Instituto de Cultura Puertorriqena (ICP). 2015. Calendario de Actividades. Accessed January 12,
2015 from http://www.icp.gobierno.pr/calendario-icp
Langellier, B.A., M.A. Martin, G. Canino, J.R. Garza, and A.N. Ortega. 2012. The Health Status of
Youth in Puerto Rico. Clinical Pediatrics 51: 569 - 573.
http://dx.doi.org/10.1177/0009922812443123
La pagina de Puerto Rico. Puerto Rico Department of Education, Puerto Rican Academic
Achievement Tests (PPAA). Accessed February 13, 2015 from
http://edfacts.dde.pr/DataElements.aspx
101
-------
Merrill, R.M., S.G. Aldana, J.E. Pope, D.R. Anderson, C.R. Coberley, J J. Grossmeier, and R.W.
Whitmer. 2013. Self-Rated Job Performance and Absenteeism According to Employee
Engagement, Health Behaviors, and Physical Health. Journal of Occupational and
Environmental Medicine 55: 10 - 18. http://dx.doi.org/10.1097/iom.0b013e31827b73af
National Oceanic and Atmospheric Administration (NOAA). 2015. National Weather Service,
Office of Climate, Water, and Weather Service, Natural Hazard Statistics. Accessed
March 24, 2015 from http://www.nws.noaa.gov/os/hazstats.shtml
Organisation for Economic Co-operation and Development (OECD). 2011. How's Life?:
Measuring well-being. Paris: OECD Publishing.
http://dx.doi.org/10.1787/9789264121164-en
Smith, L.M., H.M. Smith, J.L. Case, L.C. Harwell, J.K. Summers, and C. Wade. 2012. Indicators
and methods for constructing a U.S. Human Weil-Being Index (HWBI) for ecosystem
services research. EPA/600/R-12/023. Gulf Breeze: U.S. Environmental Protection
Agency, Office of Research and Development.
Smith, L.M., J.L. Case, L.C. Harwell, H.M. Smith, and J.K. Summers. 2013a. Development of
Relative Importance Values as Contribution Weights for Evaluating Human Wellbeing:
An Ecosystem Services Example. Human Ecology 41: 631- 641.
http://dx.doi.org/10.1007/slQ745-013-9597-5
Smith, L.M., J.L. Case, H.M. Smith, L.C. Harwell, and J.K. Summers. 2013b. Relating ecoystem
services to domains of human well-being: foundation for a U.S. index. Ecological
Indicators 28: 79-90. http://dx.doi.Org/10.1016/i.ecolind.2012.02.032
Smith, L.M., L.C. Harwell, J.K. Summers, H.M. Smith, C.M. Wade, K.R. Straub, and J.L. Case.
2014a. A U.S. Human Well-being Index (HWBI) for Multiple Scales: Linking Services
Provisioning to Human Well-being Endpoints (2000-2010). EPA/600/R-14/223.
Washington: U.S. Environmental Protection Agency.
Smith, L.M., C.M. Wade, J.L. Case, L.C. Harwell, K.R. Straub, and J.K. Summers. 2014b.
Evaluating the Transferability of a U.S. Human Weil-Being Index (HWBI) Framework to
Native American Populations. Social Indicators Research 124:157 - 182.
http://dx.doi.org/10.1007/sll205-014-Q775-7
Summers, J.K., L.M. Smith, J.L. Case, and R.A. Linthurst. 2012. A Review of the Elements of
Human Weil-Being with an Emphasis on the Contribution of Ecosystem Services. Ambio
41: 327 - 340. http://dx.doi.org/10.1007/sl3280-012-Q256-7
Summers, J.K., L.M. Smith, L.C. Harwell, J.L. Case, C.M.W.K.R. Straub, and H.M. Smith. 2014. An
index of human well-being for the U.S.: a TRIO Approach. Sustainability 6: 3915 - 3935.
http://dx.doi.org/10.3390/su6Q63915
Summers, J.K., L.C. Harwell, L.M. Smith. 2016. A model for change: An approach for forecasting
well-being from service-based decisions. Ecological Indicators 69:295-309.
Tienda, M., and F. Mitchell, (Eds.). 2006. Hispanics and the Future of America. Washington: The
National Academies Press.
102
-------
United Nations Educational, Scientific, and Cultural Organization (UNESCO). 2015. Institute for
Statistics ages 15+ 1980-2010. Accessed January 7, 2015 from
http://databank.worldbank.org/data/views/reports/metadataview.aspx
United States Census Bureau. 2012a. 2010 Urban Areas. Accessed December 9, 2015 from
http://www.census.gov/geo/reference/ua/urban-rural-2010.html
United States Census Bureau. 2012b. Household Income Inequality within U.S. Counties: 2006 -
2010. ACSBR/10-18. American Community Survey Brief.
United States Census Bureau. 2014a. Annual Estimates of the Resident Population for the
United States, Regions, States, and Puerto Rico: April 1, 2010 to July 1, 2014 (NST-
EST2014-01). Suitland: U.S. Census Bureau, Population Division.
United States Census Bureau. 2014b. Income in Puerto Rico Holds Steady After Recession.
Suitland: U.S. Census Bureau, Press Release CB14-17.
United States Census Bureau. 2015a. American Community Survey 1-Year Estimates generated
using American FactFinder. http://factfinder2.census.eov
United States Census Bureau. 2015b. Puerto Rico Public Use Microdata Sample. U.S. Census
Bureau, American Community Survey. Accessed February 3, 2015 from
http://dataferrett.census.gov
United States Department of Agriculture (USDA). 2013. Rural-Urban Continuum Codes:
Documentation. Washington: U.S. Department of Agriculture Economic Research
Services, https://www.ers.usda.gov/data-products/rural-urban-continuum-
codes/documentation.aspx#.U6wucbGtt-s
United States Department of Labor. 2015. Bureau of Labor Statistics, Mass Layoff Statistics.
Accessed June 24, 2015 from http://www.bls.eov/mIs/
United States Environmental Protection Agency (EPA). 2012. Indicators and Methods for
constructing a U.S. Human Wellbeing Index (HWBI) for ecosystem services research.
EPA/600/R-12/023. Gulf Breeze: U.S. Environmental Agency.
Villanueva, E. 2000. San Juan Bay Estuary, Comprehensive Conservation and Management Plan.
Washington: U.S. Environmental Protection Agency.
Wickham, H. 2015a. Package 'dplyr' version 0.4.1: A Grammar of Data Manipulation. R Project,
CRAN Repository, http://cran.r-proiect.ore/web/packaees/dplvr/dplyr.pdf
Wickham, H. 2015b. Package 'tidyr' version 0.2.0: Easily Tidy Data with 'spreadQ' and 'gatherQ'
Functions. R Project, CRAN Repository, http://cran.r-
proiect.ore/web/packaees/tidyr/tidyr.pdf
Wickham, H., and W. Chang. 2015. Package 'ggplot2' version 1.0.1: An Implementation of the
Grammar of Graphics. R Project, CRAN Repository.
http://cran.rproject.org/web/packages/ggplot2/ggplot2.pdf
World Bank. 2015. Life Expectancy at birth, total (years). Accessed January 27, 2015 from
http://data.worldbank.ore/indicator/SP.DYN.LEOO.IN/countries/PR-XR?display=eraph
103
-------
World Values Survey (WVS). 2015. Wave 4 1999 - 2004 OFFICIAL AGGREGATE v.20140429.
Madrid: World Values Survey Association.
Zeileis, A., G. Grothendieck, J.A. Ryan, and F. Andrews. 2015. Package 'zoo' version 1.7.12: S3
Infrastructure for Regular and Irregular Time Series (Z's Ordered Observations). R
Project, CRAN Repository, http://cran.r-proiect.org/web/packaees/zoo/zoo.pdf
Zhang, S., J. Liao, and Z. Zhu. 2008. A SAS® Macro for single imputation. Atlanta: Presented at
annual pharmaceutical industry SAS users group, June 1-4, 2008.
104
-------
Appendix A: HWBI Metrics
Table Al. Summary of metrics used for HWBI calculation for Puerto Rico, organized by
domain (bold) and indicator (italics).
Metric classification Description
Connection to Nature
Biophilia
ALLOFLFE Percentage of people who experience a connection to all of life
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: For each of the following pairs of statements, please tell me which one
comes closest to your own views (Master nature, Coexist with nature, Other answer, No
answer, Don't know). Calculated as the percentage of people who answered "Coexist with
nature"
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: General Social Survey variable ALLOFLFE, You may experience the following in
your daily life, if so how often? Experience a connection to all of life. Calculated as the
percentage of respondents who answered "Many times a day", "Every day", and "Most days"
BEAUSPRT Percentage of people who are spiritually touched by the beauty of creation
Availability: No data available
U.S. HWBI: General Social Survey BEAUSPRT, I am spiritually touched by the beauty of
creation. Calculated as the percentage of respondents who answered "Many times a day",
"Every day", and "Most days"
Cultural Fulfillment
Cultural Activity Participation
PERARTS Percentage of people who attended a jazz, classical music, opera, musical stage play, non-
musical stage play, ballet, modern, folk, and/or tap performance, or visited an art
museum/gallery or art/craft fair/festival in the past year
Source: Instituto de Cultura Puertorriqena (Instituto de Cultura Puertorriqena, 2015)
Availability: 2014-2015 (municipio)
Source measurement: Number of individual events or performances listed in Calendario de
Actividades by month Feb 2014 - Jan 2015, by 10,000 population. Calculated as (number of
cultural events)/population for each municipio. Default value is 1 if no cultural events are
advertised
Modification from U.S. HWBI: Substitute measure and units
U.S. HWBI: United States Census Bureau, American Community Survey; Census variables
PESA1A through PESA9A (Attended jazz, classical music, opera, musical stage play, non-
musical stage play, ballet, or modern, folk, tap performance, or visited an art museum/gallery
or art/craft fair/festival); Calculated as the percentage of people who responded "yes" to any
of the variables
105
-------
Metric classification
Description
Cultural Fulfillment (continued)
Cultural Activity Participation (continued)
TOTRATE All Denominations—Rates of adherence per 1000 population
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: Do you belong to a religious denomination? Calculated as the rate per
1,000 population who answered a denomination
Modification from U.S. HWBI: Modified source
U.S. HWBI: Association of Religion Data Archives, US Church Membership Data; ARDA variable
TOTRATE, All denominations/groups—Rates of adherence per 1,000 population
Education
Basic Educational Knowledge and Skills of Youth
MATHTEST Percentage of children in grades 4 and 8 with mathematics standardized test scores at or
above basic skills
Source: Puerto Rico Department of Education, Puerto Rican Academic Achievement Tests (La
pagina de Puerto Rico, 2015)
Availability: 2009 & 2010 (municipio)
Source measurement: Total students by level (pre-basic, basic, proficient, advanced) who
took the mathematics test. Calculated as the average of the percentages in grades 4 and 8 at
or above achievement level (Basic, Proficient, and Advanced) for students with complete
academic year status
Modification from U.S. HWBI: Modified source
U.S. HWBI: National Center for Education Statistics; Percentages at or above each
achievement level for mathematics, grade [4, 8] by year, jurisdiction, and All students
[TOTAL], Calculated as the average of the percentages in grades 4 and 8 at or above
achievement level
READTEST Percentage of children in grades 4 and 8 with reading standardized test scores at or above
basic skills
Source: Puerto Rico Department of Education, Puerto Rican Academic Achievement Tests (La
pagina de Puerto Rico, 2015)
Availability: 2009 & 2010 (municipio)
Source measurement: Total students by level (pre-basic, basic, proficient, advanced) who
took the Spanish reading test. Calculated as the average of the percentages in grades 4 and 8
at or above achievement level (Basic, Proficient, and Advanced) for students with complete
academic year status
Modification from U.S. HWBI: Modified source
U.S. HWBI: National Center for Education Statistics; Percentages at or above each
achievement level for reading, grade [4, 8] by year, jurisdiction, and All students [TOTAL],
Calculated as the average of the percentages in grades 4 and 8 at or above achievement level
106
-------
Metric classification
Description
Education (continued)
Basic Educational Knowledge and Skills of Youth (continued)
SCITEST Percentage of children in grades 4 and 8 with science standardized test scores at or above
basic skills
Source: Puerto Rico Department of Education, Puerto Rican Academic Achievement Tests (La
pagina de Puerto Rico, 2015)
Availability: 2009 & 2010 (municipio)
Source measurement: Total students by level (pre-basic, basic, proficient, advanced) who
took the science test. Calculated as the average of the percentages in grades 4 and 8 at or
above achievement level (Basic, Proficient, and Advanced) for students with complete
academic year status
Modification from U.S. HWBI: Modified source
U.S. HWBI: National Center for Education Statistics; Percentages at or above each
achievement level for science, grade [4, 8] by year, jurisdiction, and All students [TOTAL],
Calculated as the average of the percentages in grades 4 and 8 at or above achievement level
Participation and Attainment
ADULTLIT Percentage of people aged 16 and older who lack basic prose literacy skills
Source: United Nations Educational, Scientific, and Cultural Organization Institute for
Statistics (United Nations Educational, Scientific, and Cultural Organization, 2015)
Availability: 2004 & 2010 (Commonwealth)
Source measurement: Adult (15+) literacy rate (%). Total is the percentage of the population
age 15 and above who can, with understanding, read and write a short, simple statement on
their everyday life. Generally, 'literacy' also encompasses 'numeracy', the ability to make
simple arithmetic calculations. Calculated as 100% - literacy rate
Modification from U.S. HWBI: Modified source
U.S. HWBI: National Center for Education Statistics; Indirect estimate of percent lacking Basic
prose literacy skills and corresponding credible intervals. Percent [age 16 and older) lacking
basic prose literacy skills. Those lacking Basic prose literacy skills include those who scored
Below Basic in prose and those who could not be tested due to language barriers
HSGRAD Percentage of people aged 18 and older who obtained a high school diploma or equivalent
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015a)
Availability: 2005 - 2013 (municipio or census region)
Source measurement: Variable S1501 (Educational Attainment). Population totals and
percentages who obtained a high school (or equivalent) diploma or higher for age groups 18-
24 and 25 and older. Percentages of attainment were summed within each age group and
then averaged together using population totals as weights
Modification from U.S. HWBI: Same as U.S. HWBI
107
-------
Metric classification
Description
Education (continued)
Participation and Attainment (continued)
PARTNEDU Percentage of people aged 18 - 24 enrolled in post-secondary education
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015a)
Availability: 2005 - 2013 (municipio or census region)
Source measurement: Variable S1401 (SCHOOL ENROLLMENT) Percentage of population
enrolled in post-secondary education for population aged 18 - 24
Modification from U.S. HWBI: Modified source
U.S. HWBI: United States Census Bureau, Current Population Survey; CPS variables PETYPE-
School enrollment 2 or 4 year college, PRTAGE—single year of age and calculated as the
percentage of people aged 18 - 24 enrolled in post-secondary education
UNIVGRAD Percentage of people aged 18 and older who obtained a bachelor's degree or higher
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015a)
Availability: 2005 - 2013 (municipio or census region)
Source measurement: Variable S1501 (Educational Attainment) Population totals and
percentages who obtained a bachelor's degree or higher for age groups 18 - 24 and 25 and
older. Percentages of attainment were summed within each age group and then averaged
together using population totals as weights.
Modification from U.S. HWBI: Same as U.S. HWBI
Social, Emotional and Developmental Aspects
BULLY Percentage of children in grades 9-12 who did not go to school because they felt unsafe at
school or on their way to or from school
Source: Centers for Disease Control and Prevention, Youth Risk Behavior Surveillance System
(Centers for Disease Control and Prevention, 2015a)
Availability: 2001, 2005, & 2013 (Commonwealth)
Source measurement: Did Not Go to School Because of Safety Concerns (During the past 30
days, on how many days did you not go to school because you felt you would be unsafe at
school or on your way to or from school? Calculated as the percentage of students in grades
9-12 who had not gone to school on >1 of the 30 days preceding the survey because they
felt they would be unsafe at school or on their way to or from school)
Modification from U.S. HWBI: Same as U.S. HWBI
108
-------
Metric classification
Description
Education (continued)
Social, Emotional and Developmental Aspects (continued)
CHLDHLTH Percentage of children in excellent or very good health
Source: Asthma, Depression, and Anxiety in Puerto Rican Youth Study in Langellier et al.
(2012)
Availability: 2005 - 2008 (Commonwealth)
Source measurement: Percentage of 10 - 17 year olds who self-rated physical health as
"Excellent" (as opposed to "Very Good/Good" or "Fair/Poor")
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: United States Department of Health and Human Services, National Survey of
Children's Health; NSCH Indicator 1.1: In general, how would you describe [child name]'s
health? Would you say [his/her] health is excellent, very good, good, fair, or poor? Percentage
of children (age 0-17 years) in excellent or very good health. Note: This scale differs from
the original metric (Excellent, Very good, Good, Fair, Poor) and may underestimate
CHLDHLTH; Age limits differ: Ages 10 - 17 vs. Original metric ages 0-17
CHLDSOCIAL Percentage of children aged 6-17 years old that exhibit positive social behaviors
Source: Centers for Disease Control and Prevention, Youth Risk Behavior Surveillance System
(Centers for Disease Control and Prevention, 2015a)
Availability: 2001, 2005, & 2013 (Commonwealth)
Source measurement: In a Physical Fight (Percentage of students (9 - 12 grade) who had been
in a physical fight one or more time During the 12 months preceding the survey). Calculated
as 100% - Percentage of children in grades 9-12 who had been in a physical fight one or
more times during the 12 months preceding the survey
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: United States Department of Health and Human Services, National Survey of
Children's Health; NSCH Indicator 2.5: How many children often exhibit caring, respectful
behaviors when interacting with other children and adults? Percentage of children (age 6-17
years) who often exhibit positive social skills. "Often exhibit" is defined as answering "usually"
or "always" to at least 2 of the 4 questions [S7Q53; S7Q52; S7Q54; S7Q59],
CONFACT Percentage of respondents with a child between the ages of 3 and 5 years old, and who
recorded reading time
Availability: No data available
U.S. HWBI: Bureau of Labor Statistics, American Time Use Survey; Time spent reading to/with
household children identified by activity code 030102 (and where the youngest household
child was between the ages of 3 and 5 years old). Calculated as the percentage of parents
who have children that reported time spent (incidence, not actual time spent) reading
to/with their children
109
-------
Metric classification
Description
Health
Adults reporting having regular or personal doctor or health care provider
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2001 - 2013 (municipio, census region, or Commonwealth)
Source measurement: Variable PERSDOC2 (Do you have one person you think of as your
personal doctor or health care provider?). Calculated as the percentage of respondents who
answered "Yes"
Modification from U.S. HWBI: Same as U.S. HWBI
Average percentage of patients rating a hospital overall as 9 - 10 (1 - 10 scale)
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2013 (Commonwealth)
Source measurement: Variables CARERCVD (In general, how satisfied are you with the health
care you received?). Calculated as the percentage of respondents who answered "l=Very
satisfied"
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: Hospital Consumer Assessment of Healthcare Providers and Systems; NCAHPS
variable H_HSP_RATING_9_10, How do patients rate the hospital overall? Patients who gave a
rating of 9 or 10 (high); Calculated as the average percent of patients who gave a rating of 9
or 10
Life Expectancy and Mortality
ASTHMORT Asthma mortality as a percentage of total death
Source: Centers for Disease Control and Prevention-National Center for Health Statistics'
National Vital Statistics System Multiple Cause of Death Data (Centers for Disease Control and
Prevention, 2015e)
Availability: 2000 - 2004 (census region or Commonwealth), 2007 & 2010 - 2012 (territory)
Source measurement: Number of deaths due to asthma, age-adjusted (ICD 113 Group Code
GR113-085). Calculated as the percentage of deaths that were asthma-related
Modification from U.S. HWBI: Same as U.S. HWBI
CANCMORT Cancer mortality as a percentage of total deaths
Source: Centers for Disease Control and Prevention-National Center for Health Statistics'
National Vital Statistics System Multiple Cause of Death Data (Centers for Disease Control and
Prevention, 2015e)
Availability: 2000-2004 (municipio, census region, Commonwealth); 2007, 2010-2012
Source measurement: Number of deaths due to malignant neoplasms and various cancer
diseases, age-adjusted (ICD 113 Group Codes GR113-020 through GR113-044, excluding
GR113-037). Calculated as the percentage of deaths that were cancer-related
Modification from U.S. HWBI: Same as U.S. HWBI
Healthcare
FAMDOC
SATIS HLTHC
110
-------
Metric classification
Description
Health (continued)
Life Expectancy and Mortality (continued)
DIABMORT Diabetes mortality as a percentage of total deaths
Source: Centers for Disease Control and Prevention-National Center for Health Statistics'
National Vital Statistics System Multiple Cause of Death Data (Centers for Disease Control and
Prevention, 2015e)
Availability: 2000 - 2004 (municipio, census region, or Commonwealth), 2007 & 2010 - 2012
(territory)
Source measurement: Number of deaths due to Diabetes mellitus, age-adjusted (ICD 113
Group Code GR113-046). Calculated as the percentage of deaths that were diabetes-related
Modification from U.S. HWBI: Same as U.S. HWBI
HRTDISMORT Heart Disease mortality as a percentage of total deaths
Source: Centers for Disease Control and Prevention-National Center for Health Statistics'
National Vital Statistics System Multiple Cause of Death Data (Centers for Disease Control and
Prevention, 2015e)
Availability: 2000 - 2004 (municipio, census region, or Commonwealth), 2007 & 2010 - 2012
(territory)
Source measurement: Number of deaths due to various heart diseases and other conditions
caused by hypertension and/or high cholesterol, age-adjusted (ICD 113 Group Codes GR113-
055 through GR113-074, excluding GR113-058, -061, -064, and -072). Calculated as the
percentage of deaths that were heart disease-related
Modification from U.S. HWBI: Same as U.S. HWBI
INFMORT Infant deaths per 10,000 < 1 year of age population
Source: Centers for Disease Control and Prevention-National Center for Health Statistics'
National Vital Statistics System Multiple Cause of Death Data (Centers for Disease Control and
Prevention, 2015e)
Availability: 2000 - 2012 (municipio or Commonwealth)
Source measurement: Number of infant deaths (<1 year of age). Calculated as the number of
infant deaths per 10,000 population <1 year of age
Modification from U.S. HWBI: Same as U.S. HWBI
LIFEXPCT Life Expectancy at birth
Source: The World Bank (World Bank, 2015)
Availability: 2000-2012 (Commonwealth)
Source measurement: Life expectancy at birth (years) indicates the number of years a
newborn infant would live if prevailing patterns of mortality at the time of its birth were to
stay the same throughout its life
Modification from U.S. HWBI: Modified source
U.S. HWBI: Centers for Disease Control and Prevention; Compressed Mortality Files- all;
Calculated using CDC's Compressed Mortality Files and Fergany's (1971) methods. Life
expectancy was determined by county-level age group rates; missing or zero age group rates
were imputed from the next higher spatial level (state or national)
111
-------
Metric classification
Description
Health (continued)
Life Expectancy and Mortality (continued)
SUICDMORT Suicide mortality as a percentage of total deaths
Source: Centers for Disease Control and Prevention-National Center for Health Statistics'
National Vital Statistics System Multiple Cause of Death Data (Centers for Disease Control and
Prevention, 2015e)
Availability: 2000 - 2004 (census region or Commonwealth), 2007 & 2010 - 2012 (territory)
Source measurement: Number of deaths due to intentional self-harm, age-adjusted (ICD 113
Group Codes GR113-125 and GR113-126). Calculated as the percentage of deaths that were
suicide-related
Modification from U.S. HWBI: Same as U.S. HWBI
Lifestyle and Behavior
ALCOHOL Percentage of population drinking on average more than 1 drink per day
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2000 - 2012 (municipio, census region, or Commonwealth)
Source measurement: CDC BRFSS variables DRINKANY, DRNKANY2, DRNKANY3, and
DRNKANY4, During the past month have you had at least one drink of any alcoholic beverage
such as beer, wine, wine coolers, or liquor? 2) CDC variables ALCDAYS, ALCDAY3, ALCDAY4,
and ALCOHOL, During the past 30 days, how many days per week or per month did you have
at least one drink of any alcoholic beverage? 3) CDC variables NALCOCC, AVEDRNK, and
AVEDRNK2, On the days when you drank, about how many drinks did you drink on the
average? Calculated as the percentage of people who drank on average more than one drink
per day using the variables listed above
Modification from U.S. HWBI: Same as U.S. HWBI
HBI Healthy Behaviors Index
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2000, 2003, 2005, 2007, 2009, 2011 (municipio, census region, or
Commonwealth)
Source measurement: The Healthy Behaviors Index (HBI) is a mean of 3 items recoded to
reflect the positive responses only. CDC BRFSS variables RFPAMOD, RFREGUL, TOTINDB (Risk
factor for moderate physical activity defined as 30 or more minutes per day for 5 or more
days per week, or vigorous activity for 20 or more minutes per day on 3 or more days),
FRTINDEX (summary index based on the calculated number of daily servings of fruits and
vegetables), and SMOKER2 and SMOKER3 (Four level smoker status: Every day smoker,
Someday smoker, Former smoker, Non-smoker). HBI calculation described in Merrill et
al.(2013); HBI= ((HBI RFPAMOD + HBI FRTINDEX + HBI_RFSMOK3)/3)*100
112
-------
Metric classification
Description
Health (continued)
Lifestyle and Behavior (continued)
HBI (continued)
Modification from U.S. HWBI: Alternative U.S. HWBI
U.S. HWBI: Gallup-Healthways; The Healthy Behaviors Index (HBI) is a mean of four items
recoded to reflect the positive responses only. The four items are Gallup variables Hll (Do
you smoke?), M16 (Did you eat healthy all day yesterday?), H12A (if respondent reported
exercising 3-7 times per week), and H12B (if respondent reported eating 5 fruits and
vegetables per day, 4 or more times per week). Alternate Source: CDC BRFSS variables
RFPAMOD (Risk factor for moderate physical activity defined as 30 or more minutes per day
for 5 or more days per week, or vigorous activity for 20 or more minutes per day on 3 or more
days), FRTINDEX (summary index based on the calculated number of daily servings of fruits
and vegetables), and SMOKER2 and SMOKER3 (Four level smoker status: Every day smoker,
Someday smoker, Former smoker, Non-smoker); The average index value was calculated for
each county (Gallup). The average of the variables was computed at the respondent level
following the same recoding procedure as Gallup (CDC)
TEENPREG Births to mothers under 20 years old (as percentage of live births)
Source: Centers for Disease Control and Prevention, National Center for Health Statistics
(NCHS), National Vital Statistics System, Birth Data Files (Centers for Disease Control and
Prevention, 2015d)
Availability: 2000 - 2012 (municipio or Commonwealth)
Source measurement: CDC variables for year, county of residence, and age of mother.
Calculated as the percentage of births to mothers in the age groups "under 15" and "15 - 19"
as a percentage of all births
Modification from U.S. HWBI: Same as U.S. HWBI
TEENSMK Percentage of population (grades 9 - 12) reporting they Smoked Cigarettes On 20 Or More
Days in previous 30 days
Source: Centers for Disease Control and Prevention, Youth Risk Behavior Surveillance System
(Centers for Disease Control and Prevention, 2015a)
Availability: 2005, 2011, 2013 (Commonwealth)
Source measurement: Current Frequent Cigarette Use (Percentage of children in grades 9 -
12 who smoked cigarettes on 20 or more days in the past 30 days)
Modification from U.S. HWBI: Same as U.S. HWBI
113
-------
Metric classification
Description
Health (continued)
Personal Well-being
HAPPY Percentage of people who are very happy or pretty happy
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey )
Source measurement: Taking all things together, would you say you are: (Very happy, Quite
happy, Not very happy, Not at all happy, No answer, Don't know). Calculated as the
percentage of people who answered "Very happy" and "Quite happy"
Modification from U.S. HWBI: Modified source
U.S. HWBI: General Social Survey; GSS variable HAPPY, Taken all together, how would you say
things are these days- Would you say that you are very happy, pretty happy, or not too
happy? Alternate Source: Gallup Healthways variable WP6878, Did you experience happiness
a lot of the day yesterday? Calculated as the percentage of respondents who answered "Very
happy" or "Pretty happy" (GSS); and the percentage of respondents who answered "Yes"
(Gallup)
LIFESATIS Adults reporting satisfied with life
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2005 - 2010 (municipio, census region, or Commonwealth)
Source measurement: CDC BRFSS variable LSATISFY (2005 - 2010); In general how satisfied
are you with your life? Calculated as the proportion of people who are "Very satisfied" or
"Satisfied" with their life
Modification from U.S. HWBI: Same as U.S. HWBI
PRCVDHLTH Adults reporting general health is good, very good, or excellent
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2000 - 2012 (municipio, census region, or Commonwealth)
Source measurement: CDC BRFSS variable GENHLTH (2000 - 2013), Would you say that in
general your health is: excellent, very good, good, fair, or poor? Calculated as the percentage
of people who responded that their health was "Excellent", "Very Good" or "Good"
Modification from U.S. HWBI: Same as U.S. HWBI
Physical and Mental Health Conditions
ADLTASTHMA Adults reporting diagnosed with asthma in lifetime
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2000 - 2013 (municipio, census region, or Commonwealth)
Source measurement: Variables ASTHMA (2000), ASTHMA2 (2001 - 2010), and ASTHMA3
(2011 - 2013) Has a doctor or other health professional ever told you that you had asthma?
Calculated as the percentage of respondents who answered "Yes"
Modification from U.S. HWBI: Same as U.S. HWBI
114
-------
Metric classification
Description
Health (continued)
Physical and Mental Health Conditions (continued)
CANCER Adults reporting diagnosed with cancer in lifetime
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2009, 2011 - 2013 (municipio, census region, or Commonwealth)
Source measurement: Variables CNCRHAVE (2009), CHCSCNCR + CHCOCNCR (2011 - 2013);
Have you EVER been told by a doctor, nurse, or other health professional that you had
cancer? Calculated as the percentage of people who responded "Yes"
Modification from U.S. HWBI: Same as U.S. HWBI
CHLDASTHMA Adults reporting 1 or more child in HH was diagnosed with asthma in lifetime
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2005 - 2013 (municipio, census region, or Commonwealth)
Source measurement: Variable CASTHDX2 (2005 - 2013); Earlier you said there were [fill in
number from core Q12.6] children age 17 or younger living in your household. How many of
these children have ever been diagnosed with asthma? Calculated as the percentage of
respondents who reported 1 or more child
Modification from U.S. HWBI: Same as U.S. HWBI
DEPRESSION Adults reporting diagnosed with depression in lifetime
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2011 & 2013 (municipio, census region, or Commonwealth)
Source measurement: Variable ADDEPEV (2006; 2010), ADDEPEV2 (2011 - 2013); Has a
doctor or other healthcare provider ever told you that you have a depressive disorder
(including depression, major depression, dysthymia, or minor depression)? Calculated as the
percentage of respondents who answered "Yes"
Modification from U.S. HWBI: Same as U.S. HWBI
DIABETES Adults diagnosed with diabetes in lifetime
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2000 - 2013 (municipio, census region, or Commonwealth)
Source measurement: Variables DIABETES (2000 - 2003), DIABETE2 (2004 - 2010), DIABETE3
(2011 - 2013); Have you ever been told by a doctor that you have diabetes? (If "Yes" and
respondent is female, ask "Was this only when you were pregnant?"). Calculated as the
percentage of people who responded "Yes"
Modification from U.S. HWBI: Same as U.S. HWBI
115
-------
Metric classification
Description
Health (continued)
Physical and Mental Health Conditions (continued)
HRTATTACK Adults reporting diagnosed with heart attack or myocardial infarction in lifetime
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2003 & 2005 - 2013 (municipio, census region, or Commonwealth)
Source measurement: Variables CVDINFR2 (2002 - 2003), CVDINFR3 (2005 - 2006), CVDINFR4
(2007 - 2013); Has a doctor, nurse, or other health professional ever told you that you had
any of the following? A heart attack, also called a myocardial infraction. Calculated as the
percentage of respondents who answered "Yes"
Modification from U.S. HWBI: Same as U.S. HWBI
HRTDISEASE Adults reporting diagnosed with angina or coronary heart disease in lifetime
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2003 & 2005 - 2013 (municipio, census region, or Commonwealth)
Source measurement: Variables CVDCRHD2 (2002 - 2003), CVDCRHD3 (2005 - 2006),
CVDCRHD4 (2007 - 2013); [Has a doctor, nurse or other health professional ever told you that
you had any of the following.] Angina or coronary heart disease. Calculated as the percentage
of respondents who answered "Yes"
Modification from U.S. HWBI: Same as U.S. HWBI
OBESITY Age adjusted percentage of population 18+ classified as obese (BMI >=30)
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2000 - 2013 (municipio, census region, or Commonwealth)
Source measurement: Variables BMI2CAT (2000 - 2002), BMI3CAT (2003), BMI4CAT (2004 -
2010), BMI5CAT (2011 - 2013); Calculated categories of Body Mass Index: Neither
overweight nor obese, Obese, Overweight. Calculated as the percentage of respondents
categorized as obese (BMI>=30)
Modification from U.S. HWBI: Modified source
U.S. HWBI: Centers for Disease Control and Prevention, National Diabetes Surveillance
System; NDSS variable ADJPERCENT, age-adjusted percentage of population aged 18 and
older classified as obese (BMI>=30)
STROKE Adults reporting diagnosed with stroke in lifetime
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2003; 2005-2012 (municipio, census region, or Commonwealth)
Source measurement: Variables CVDSTRK2 (2003), CVDSTRK3 (2005 - 2013); [Has a doctor,
nurse or other health professional ever told you that you had any of the following.] A stroke.
Calculated as the percentage of respondents who answered "Yes"
Modification from U.S. HWBI: Same as U.S. HWBI
116
-------
Metric classification
Description
Leisure Time
Leisure Activity Participation
PHYSACTIV Percentage of adults (aged 18 years and older) who participated in physical activities or
exercises in the past 30 days
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2000 - 2013 (municipio, census region, or Commonwealth)
Source measurement: Variables EXERANY (2000), EXERANY2 (2001 - 2013); During the past
month, did you participate in any physical activities or exercises such as running, calisthenics,
golf, gardening, or walking for exercise? Calculated as the percentage of people who
answered "yes"
Modification from U.S. HWBI: Same as U.S. HWBI
VACATION Average number of nights away from home on vacation or visiting friends/relatives
Source: Tourism Company of Puerto Rico (PRTC), Market Studies (reported by Instituto de
Estadisticas de Puerto Rico) (Compania de Turismo de Puerto Rico, 2015)
Availability: 2007-2013 (Commonwealth)
Source measurement: Displays the number of visitors registered in hotels endorsed by the
Tourism Company (Calendar Year), by geographic region, by country and in the case of the
United States by state (hotel nights by geography (PR residents)). Calculated as the number of
resident visitors as a percentage of the total population
Modification from U.S. HWBI: Substitute measure and units
U.S. HWBI: Bureau of Labor Statistics, American Time Use Survey, Trips Survey Supplement;
BLS variable TUTRV2- Main purpose for the trip, and BLS variable TUTRV5-Total nights away
from home; Calculated as the average number of nights away from home when the main
purpose was vacation or visiting friends/relatives
Time Spent
LEISURE Average number of minutes spent on relaxing and leisure by adults
Source: United Nations Research Institute for Social Development, Gender and Development
Programme (Budlender, 2008)
Availability: 2000 (substitute region)
Source measurement: Non-productive time (not paid or unpaid work). Calculated as average
Non-productive time of Nicaragua (1998) and Buenos Aires (2005) - sleep hours (=8hrs (=480
minutes))
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: Bureau of Labor Statistics; Time spent on socializing, relaxing, leisure and sports
identified by activity codes 12xxxx-13xxxx (where "xx" indicates any numbers to complete the
6-digit activity code from the coding lexicon). Calculated as the average percentage of time
involved in these activities
117
-------
Metric classification
Description
Leisure Time (continued)
Working Age Adults
LONGWRKHRS Population reporting that they were working 50 hours per week or more
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015b)
Availability: 2005 - 2007 & 2009 - 2013 (census region)
Source measurement: Percentage of employed respondents reporting that they work 50
hours or more per week
Modification from U.S. HWBI: Modified source
U.S. HWBI: Bureau of Labor Statistics and US Census Bureau joint effort CPS variable
PEHRUSLT, hours usually worked at all jobs
NORMWRKHRS Proportion of work activity (act code: 0501xx) during daytime (9 - 5) hours to total work
activity duration
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015b)
Availability: 2005 - 2007 & 2009 - 2013 (census region)
Source measurement: Percentage of work activity duration during daytime hours (9 am to 5
pm) from total work activity duration. Calculated as average hours worked per day intersect
with interval 9am - 5pm (where average hours worked per day are average hours per week/5
days)
Modification from U.S. HWBI: Modified source
U.S. HWBI: Bureau of Labor Statistics, American Time Use Survey Work and work-related
activities identified by activity codes 0501xx (where "xx" indicates any numbers to complete
the 6-digit activity code from the coding lexicon)
SENIORCARE Average number of minutes spent on adult care activities (activity codes 0304xx, 0305xx,
0404xx, 0405xx)
Availability: No data available
U.S. HWBI: Bureau of Labor Statistics, American Time Use Survey; Adult care activities
identified by activity codes 0304xx, 0305xx, 0404xx, 0405xx (where "xx" indicates any
numbers to complete the 6-digit activity code from the coding lexicon). Calculated as the
percentage of adult care activities duration from total activities duration
118
-------
Metric classification
Description
Living Standards
Basic Necessities
FOODSECURE Percent of households that had high or marginal food security
Source: Gallup World Poll 2006 in Gasparini et al., (2010)
Availability: 2006 (Commonwealth)
Source measurement: "Have there been times in the past twelve months when you did not
have enough money to buy food that you or your family needed?" Calculated as percentage
of population with enough money to buy food
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: United States Census Bureau, Current Population Survey; Census variable
HRFS12M1, Food Security Summary Status, 12-month; Calculated as the percentage of
households that responded "Food Secure - High or Marginal Food Security"
HOMEAFFORD Median selected monthly owner costs as a percentage of household income
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015a)
Availability: 2005 - 2013 (municipio and census region)
Source measurement: Variable B25092, Median selected monthly owner costs as a
percentage of household income, Total
Modification from U.S. HWBI: Same as U.S. HWBI
Median household income
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015a)
Availability: 2005 - 2013 (municipio or census region)
Source measurement: Variable B19013 Median household income in the past 12 months
(inflation-adjusted dollars)
Modification from U.S. HWBI: Modified source
U.S. HWBI: United States Census Bureau, Small Area Income and Poverty Estimates Median
household income, in dollars; number
POVERTY Percentage of population (all ages) in poverty
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015a)
Availability: 2005 - 2013 (municipio or census region)
Source measurement: Variable S1701 (POVERTY STATUS IN THE PAST 12 MONTHS).
Calculated as the percent below poverty level for population for whom poverty status is
determined
Modification from U.S. HWBI: Modified source
U.S. HWBI: United States Census Bureau, Small Area Income and Poverty Estimates All ages in
poverty; Percent
Income
MEDINCOME
119
-------
Metric classification
Description
Living Standards (continued)
Income (continued)
POVPERSIST Percentage of respondents who are currently in poverty and stated that their financial
situation has remained the same over the past few years
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015a)
Availability: 2005 - 2013 (municipio or census region)
Source measurement: Percentage of Population in poverty (Poverty status*Population)
compared to previous year (2006 - 2013). Calculated as the percentage of population in
poverty compared to previous year (Population in poverty in Year B/Population in Poverty in
Year A)* 100%
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: General Social Survey; 1) U.S. Census Bureau weighted average poverty threshold
for the year 1986. 2) GSS variable REALINC: Family income on 1972 - 2006 surveys in constant
dollars (base = 1986). 3) GSS variable FINALTER: During the last few years, has your financial
situation been getting better, worse, or has it stayed the same? 4) GSS variable HOMPOP
Household Size and Composition; Calculated as the percentage of respondents who answered
"Stayed the same" for GSS variable FINALTER, while using the responses to GSS variables
REALINC and HOMPOP to determine what respondents were below the U.S. Census poverty
thresholds
Wealth
HOMEVAL Median value of owner occupied housing units
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015a)
Availability: 2005 - 2013 (municipio or census region)
Source measurement: Variable B25077, Median value of owner-occupied housing units
Modification from U.S. HWBI: Same as U.S. HWBI
MTGDEBT Percentage of owner-occupied housing units without a second mortgage or home equity loan
Source: United States Census Bureau, American Community Survey (United States Census
Bureau, 2015a)
Availability: 2005 - 2013 (municipio or census region)
Source measurement: Variable B25081, Mortgage status of owner-occupied housing units.
Calculated as the percentage of owner-occupied housing units with no second mortgage or
home equity loan
Modification from U.S. HWBI: Same as U.S. HWBI
120
-------
Metric classification
Description
Living Standards (continued)
Work
JOBLOSE Percentage of people that responded it is "not too likely" or "not at all likely" that they will
lose their job or be laid off
Source: United States Department of Labor, Bureau of Labor Statistics Mass Layoff Statistics
(2000 - 2012) (United States Department of Labor, 2015)
Availability: 2000-2012 (Commonwealth)
Source measurement: Total initial claimants (The total number of initial claimants associated
with the mass layoffs and/or extended mass layoffs. "Initial claimant" is a term used to define
the person who files an initial notice of unemployment with the State Unemployment
Insurance agency for either for a determination of entitlement to and eligibility for
compensation, or for a subsequent period of unemployment within a benefit year or period
of eligibility. Mass Layoffs - situations that involve establishments which have at least 50
initial claims for unemployment insurance (Ul) filed against them during a consecutive 5-week
period). Calculated as the percentage of the labor force that did not file as an initial claimant
associated with a mass layoff
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: General Social Survey; GSS variable JOBLOSE, Thinking about the next 12 months,
how likely do you think it is that you will lose your job or be laid off- very likely, fairly likely,
not too likely, or not at all likely? Calculated as the percentage of respondents who answered
"Not too likely" or "Not at all likely"
JOBSATIS Percentage of people who are satisfied with their job or the work they do
Source: Kelly Services (2013 Kelly Global Workforce Index) in Caribbean Business (2013)
Availability: 2013 (Commonwealth)
Source measurement: Percentage of respondents who feel totally committed to their current
employer
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: Gallup-Healthways; GSS variable SATJOB1, All in all how satisfied would you say
you are with your job? Alternate Source: Gallup Healthways variable WP9045, Are you
satisfied or dissatisfied with your job or the work you do? Calculated as the percentage of
respondents who answered "Very Satisfied" and "Somewhat Satisfied" (GSS), and the
percentage of respondents who answered "Satisfied" (Gallup)
121
-------
Metric classification
Description
Safety and Security
Actual Safety
ACCMM Total reported number of accidental morbidity and mortality cases excluding weather events
(See NATHAZLOSS)
Source: Centers for Disease Control and Prevention-National Center for Health Statistics'
National Vital Statistics System Multiple Cause of Death Data (Centers for Disease Control and
Prevention, 2015e)
Availability: 2000 - 2007 & 2010 - 2012 (municipio, census region, or Commonwealth)
Source measurement: Number of deaths due to external causes (ICD-10 Group Codes V01
through Y89), excluding deaths caused by natural hazards and intentional deaths (ICD-10
group codes X30-X39, X60-X84, Y85-Y89). Calculated as the number of deaths per 100,000
population that were accident-related
Modification from U.S. HWBI: Same as U.S. HWBI
NATHAZHLOSS Total # of injuries and fatalities attributed to natural events
Source: National Oceanic and Atmospheric Administration (NOAA) National Weather Service
Natural Hazard Statistics (National Oceanic and Atmospheric Administration, 2015)
Availability: 2000-2013 (Commonwealth)
Source measurement: Fatalities and injuries by state for cold, flood, heat, lightning, tornado,
tropical cyclone, wind, and winter storm. Calculated as the number of fatalities and injuries
from hazardous weather per 100,000 population
Modification from U.S. HWBI: Modified source
U.S. HWBI: University of South Carolina, Hazards and Vulnerability Research Institute, Spatial
Hazard Events and Losses Database for the United States (SHELDUS); SHELDUS dataset,
Fatalities and injuries from hazardous weather; Calculated as the number of fatalities and
injuries from hazardous weather per 100,000 population
PROPCRIME Number of property crimes per 100,000 people
Source: Federal Bureau of Investigation (FBI) Uniform Crime Reports (Federal Bureau of
Investigation, 2015)
Availability: 2000-2013 (census region)
Source measurement: Number of crimes. Calculated as the total (sum) number of property
crimes (Burglary, Larceny theft, Motor vehicle theft, Arson) per 100,000 people. Population
estimates reflect the total population served by reporting agencies
Modification from U.S. HWBI: Modified source
U.S. HWBI: National Archives of Criminal Justice Data; NACJD variables BURGLRY, LARCENY,
MVTHEFT, ARSON, Number of burglary, larceny, motor vehicle theft, and arson offenses;
Calculated as the total (sum) number of property crimes per 100,000 people. Population
estimates were provided by the NACJD (variable CPOPCRIM) and reflect the total population
served by reporting agencies
122
-------
Metric classification
Description
Safety and Security (continued)
Actual Safety (continued)
VIOLCRIME Number of violent crimes per 100,000 people
Source: Federal Bureau of Investigation (FBI) Uniform Crime Reports (Federal Bureau of
Investigation, 2015)
Availability: 2000-2013 (census region)
Source measurement: Number of crimes. Calculated as the total (sum) number of violent
crimes (Murder and non-negligent manslaughter, Forcible rape, Robbery, Aggravated assault)
per 100,000 people. Population estimates reflect the total population served by reporting
agencies
Modification from U.S. HWBI: Modified source
U.S. HWBI: National Archives of Criminal Justice Data; NACJD variables MURDER, RAPE,
ROBBERY, AGASSLT, Number of murder, rape, robbery, and aggravated assault offenses;
Calculated as the total (sum) number of violent crimes per 100,000 people. Population
estimates were provided by the NACJD (variable CPOPCRIM) and reflect the total population
served by reporting agencies
Perceived Safety
PRCVDSAFE Percentage of people who feel safe walking alone at night where they live
Source: Gaither International in Caribbean Business (2014a)
Availability: 2013 (Commonwealth)
Source measurement: Percentage of population feeling very comfortable or comfortable
walking alone in their neighborhoods at night
Modification from U.S. HWBI: Modified source
U.S. HWBI: Gallup-Healthways; Gallup variable WP113, Do you feel safe walking alone at night
in the city or area where you live? Calculated as the percentage of people who responded
"Yes"
Risk
SOVI Social Vulnerability Index (SoVI) for the United States
Source: Indices of Social Vulnerability to Natural Hazards: A Comparative Evaluation (Gall,
2007)
Availability: 2000 (Commonwealth)
Source measurement: PIV (Predictive Indicators of Vulnerability). Index Score based on Adger
et al., 2004
Modification from U.S. HWBI: Substitute measure and units
U.S. HWBI: University of South Carolina, Hazards and Vulnerability Research Institute, Spatial
Hazard Events and Losses Database for the United States (SHELDUS); Social Vulnerability
Index (SoVI®) for the United States, SoVI Score. This index estimates a population's ability to
prepare for, respond to, and recover from environmental hazards. Higher scores indicate
more vulnerability
123
-------
Metric classification
Description
Social Cohesion
Attitude toward Others and the Community
CANTRUST Percentage of people who think that others can almost always or can usually be trusted
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: Generally speaking, would you say that most people can be trusted or
that you need to be very careful in dealing with people? (Most people can be trusted, Need to
be very careful, No answer, Don't know). Calculated as the percentage of people who
answered "Most people can be trusted"
Modification from U.S. HWBI: Modified source
U.S. HWBI: General Social Survey; GSS variable CANTRUST, Generally speaking, would you say
that people can be trusted or that you can't be too careful in dealing with people? Calculated
as the percentage of respondents who answered "people can almost always be trusted" and
"people can usually be trusted"
CITYSATIS Percentage of people who are satisfied with the city or area where they live
Source: Gaither International in Caribbean Business (2014b)
Availability: 2011, 2013, 2014 (Commonwealth)
Source measurement: Percentage of population considering leaving Puerto Rico. Calculated
as 100% - the percentage of respondents who are considering moving out of Puerto Rico
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: Gallup-Healthways; Gallup variable WP83, Are you satisfied or dissatisfied with the
city or area where you live? Calculated as the percentage of respondents who answered
"Satisfied"
CLSETOWN Percentage of people who feel close to their town or city
Source: Gaither International in Caribbean Business (2011)
Availability: 2011 (Commonwealth)
Source measurement: Percentage of people who have good relationships with their
neighbors. Calculated as percent of population having very good or good relationship with
neighbors
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: General Social Survey; GSS variable CLSETOWN, How close do you feel to your
town or city? Calculated as the percentage of respondents who answered "Very Close" and
"Close"
DISCRM2 Percentage of people that were emotionally upset as a result of how they were treated based
on their race
Availability: No data available
U.S. HWBI: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System; CDC Variable RREMTSM1, Within the past 12 months on average, how often have
you felt emotionally upset, for example angry, sad, or frustrated, as a result of how you were
treated based on your race? Calculated as the percentage of respondents who answered
anything except "Never"
124
-------
Metric classification
Description
Social Cohesion (continued)
Attitude toward Others and the Community (continued)
HELPFUL Percentage of people who think that others try to be helpful
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: Do you think most people would try to take advantage of you if they
got a chance, or would they try to be fair? (Would take advantage, Try to be fair, No answer,
Don't know). Calculated as the percentage of people who answered "Try to be fair"
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: General Social Survey; GSS variable HELPFUL, Would you say that most of the time
people try to be helpful, or that they are mostly just looking out for themselves? Calculated as
the percentage of people who responded "Try to be helpful"
Democratic Engagement
POLINTRST Percentage of people interested in political campaigns
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: How interested would you say you are in politics (Very interested,
Somewhat interested, Not very interested, Not at all interested, No answer, Don't know)?
Calculated as the percentage of people who answered "Somewhat interested" or "Very
interested"
Modification from U.S. HWBI: Modified source
U.S. HWBI: American National Election Studies; ANES variable VCF0310, Some people don't
pay much attention to political campaigns. How about you, would you say that you have
been/were very much interested, somewhat interested, or not much interested in the
political campaigns (so far) this year? Alternate Source: General Social Survey (GSS) variable
POLINT and POLINT1, How interested would you say you personally are in politics? Calculated
as the percentage of people who answered "Somewhat interested" or "Very much
interested" for variable VCF0310 (ANES). Calculated as the percentage of people who
answered "Very interested", "Fairly interested", or "Somewhat interested" for variable
POLINT, and the percentage of people who answered "Very interested" or "Fairly interested"
for variable POLINT1 (GSS)
REGVOTRS Percentage of U.S. citizens (aged 18 and older) that are registered to vote
Source: Comision Estatal de Elecciones (Comision Estatal de Elecciones, 2015)
Availability: 2004, 2008, 2012 (municipio)
Source measurement: Calculated as the percentage of registered voters from population ages
18 and over
Modification from U.S. HWBI:
U.S. HWBI: United States Census Bureau, Current Population Survey Percentage of U.S.
citizens aged 18 and older (eligible voters) that are registered to vote
125
-------
Metric classification
Description
Social Cohesion (continued)
Democratic Engagement (continued)
SATDEM Percentage of people who are satisfied with democracy in the United States
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: On the whole are you very satisfied, rather satisfied, not very satisfied
or not at all satisfied with the way democracy is developing in our country? (Very satisfied,
Rather satisfied, Not very satisfied, Not at all satisfied, No answer, Don't know). Calculated as
the percentage of people who answered "Rather satisfied" or "Very satisfied"
Modification from U.S. HWBI: Modified source
U.S. HWBI: General Social Survey; GSS Variable DEMTODAY, How well does democracy work
in America today? On the whole, on a scale of 0 to 10 where 0 is very poorly and 10 is very
well. GSS Variable SATDEMOC, On the whole, are you very satisfied, fairly satisfied, not very
satisfied, or not at all satisfied with the way democracy works in the United States? Calculated
as the percentage of respondents who answered 6 through 10 for the variable DEMTODAY,
and "very satisfied" and "fairly satisfied" for the variable SATDEMOC
TRUSTGOV Percentage of people who trust the government to do what is right
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: I am going to name a number of organizations. For each one, could you
tell me how much confidence you have in them: is it a great deal of confidence, quite a lot of
confidence, not very much confidence or none at all? The government in [WASHINGTON/
YOUR CAPITAL] (A great deal, Quite a lot, Not very much, None at all, No answer, Don't
know). Calculated as the percentage of people who answered "Quite a lot" or "A great deal"
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: American National Election Studies; ANES Variable VCF0604, People have different
ideas about the government in Washington. These ideas don't refer to Democrats or
Republicans in particular, but just government in general. We want to see how you feel about
these ideas. How much of the time do you think you can trust the government in Washington
to do what is right - just about always, most of the time, only some of the time? Alternate
Source: General Social Survey (GSS) variable POLEFF17, Most government administrators can
be trusted to do what is best for the country. Calculated as the percentage of respondents
who answered "Most of the time" or "Just about always" for the variable VCF0604 (ANES),
and the percentage of respondents who answered "Strongly agree" or "Agree" for the
variable POLEFF17 (GSS)
126
-------
Metric classification
Description
Social Cohesion (continued)
Democratic Engagement (continued)
VOICENGOV Percentage of people who feel that public officials care about what they think
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: Generally speaking, would you say that this country is run by a few big
interests looking out for themselves, or that it is run for the benefit of all the people? (Run by
few big interests, Run for all people, No answer, Don't know). Calculated as the percentage of
people who answered "Run for all people"
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: American National Election Studies; ANES Variable VCF0609, Please tell me how
much you agree or disagree with this statement: Public officials don't care much what people
like me think. Alternate Source: General Social Survey (GSS) variable POLEFF11, People like
me don't have any say about what the government does. Calculation Methods: Calculated as
the percentage of respondents who answered "Disagree" for variable VCF0609 (ANES), and
the percentage of respondents who answered "Disagree" or "Strongly disagree" for the
variable POLEFF11 (GSS)
VOTRTOUT Percentage of U.S. citizens (aged 18 and older) that voted
Source: Comision Estatal de Elecciones (Comision Estatal de Elecciones, 2015)
Availability: 2004, 2008, 2012 (municipio)
Source measurement: Calculated as the percentage of population ages 18 and over who
voted
Modification from U.S. HWBI: Modified source
U.S. HWBI: United States Census Bureau, Current Population Survey Percentage of U.S.
citizens aged 18 and older (eligible voters) Percentage of U.S. citizens (aged 18 and older) that
voted
Average number of minutes parents spend reading to children
Availability: No data available
U.S. HWBI: Bureau of Labor Statistics, American Time Use Survey; Adults reading to children
identified by activity codes 030102 and 040102. Calculated as the percentage of parent-child
reading activity duration from total activities duration
MEALS Percentage of time spent by children (aged 15 - 17 years old) eating at home with parents
from total eating time
Availability: No data available
U.S. HWBI: Bureau of Labor Statistics, American Time Use Survey; Time spent by children,
aged 15 - 17 years old, eating at home with their parents, identified by activity codes llxxxx
(where "xx" indicates any numbers to complete the 6-digit activity code from the coding
lexicon). Calculated as the percentage of time spent eating at home with parents by children
(aged 15 - 17) from the child's total eating time
Family Bonding
CHLDREAD
127
-------
Metric classification
Description
Social Cohesion (continued)
Family Bonding
WATCHTV Percentage of children (9 - 12 grade) Watching Television 3 Or More Hours Per Day (on an
average school day)
Source: Centers for Disease Control and Prevention, Youth Risk Behavior Surveillance System
(Centers for Disease Control and Prevention, 2015a)
Availability: 2005, 2011, 2013 (Commonwealth)
Source measurement: Watched Television >3 Hours/Day. Calculated as percentage of
students grades 9-12 who watched television >3 hours/day on an average school day
Modification from U.S. HWBI: Same as U.S. HWBI
Social Engagement
CHLDACTV Percentage of children who participate in one or more organized activities outside of school,
such as sports teams or lessons, clubs, or religious groups
Source: Centers for Disease Control and Prevention, Youth Risk Behavior Surveillance System
(Centers for Disease Control and Prevention, 2015a)
Availability: 2005, 2011, 2013 (Commonwealth)
Source measurement: Played on >1 Sports Teams (Calculated as the percentage of students
grades 9-12 who had played on >1 sports teams (run by their school or community groups)
during the 12 months preceding the survey)
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: United States Department of Health and Human Services, National Survey of
Children's Health; Percentage of children aged 6-17 years old who participate in one or
more organized activities outside of school
GRPACTV Percentage of people who are a member of any type of organization
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: Please look carefully at the following list of voluntary organizations and
activities and say...which, if any, do you belong to? Calculated as the percentage of people
who answered "Belong" for at least one organization (social welfare service for elderly,
church organization, cultural activities, labor unions, political parties, local political, human
rights, animal rights, professional associations, youth work, sports or recreation, women's
groups, peace movement, concerned with health, other groups)
Modification from U.S. HWBI: Modified source
U.S. HWBI: General Social Survey; GSS variable MEMNUM, Could you tell me whether or not
you are a member of any type of organization? Calculated as the percentage of people who
are members of one or more groups
128
-------
Metric classification
Description
Social Cohesion (continued)
Social Engagement (continued)
VOLNTR Volunteer Rate (percentage of Current Population Survey respondents who volunteered)
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: Please look carefully at the following list of voluntary organizations and
activities and say...which, if any, are you currently doing unpaid voluntary work? Calculated as
the percentage of people who answered "Do voluntary work" for at least one volunteer
activity (social welfare, church organization, cultural activities, labor unions, political parties,
local community action, third world human, conservation, professional associations, youth
work, sports or recreation, women's groups, peace movement, concerned with health, other
social groups)
Modification from U.S. HWBI: Modified source
U.S. HWBI: Bureau of Labor Statistics and US Census Bureau joint effort; Volunteer rate
(equals the percentage of Current Population Survey respondents who reported that they had
performed any unpaid volunteer work)
Social Support
CLSFRNDFAM Percentage of people who have 6 or more close friends and/or relatives
Source: World Values Survey (World Values Survey, 2015)
Availability: 2001 (World Values Survey Region)
Source measurement: For each of the following, indicate how important it is in your life.
Would you say it is: v4=Family; v5=Friends (Very important, Rather important, Not very
important, Not at all important, No answer, Don't know). Calculated as the percentage of
people who answered "Very Important"
Modification from U.S. HWBI: Substitute measure
U.S. HWBI: General Social Survey; GSS variable NUMPROBS, Of these (NUMCNTCT) friends
and relatives, about how many would you say you feel really close to, that is close enough to
discuss personal or important problems with? (variable NUMCNTCT: Not counting people at
work or family at home, about how many other friends or relatives do you keep in contact
with at least once a year?). Calculated as the percentage of respondents who answered 6 or
more friends or relatives
EMTSUPRT Proportion of participants responding that they usually or always get the emotional and social
support they need
Source: Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
System (Centers for Disease Control and Prevention, 2015b)
Availability: 2005 - 2010 (municipio, census region, or Commonwealth)
Source measurement: BRFSS variable EMTSUPRT (2005 - 2010), How often do you get the
social and emotional support you need? Calculated as the percentage of people who
responded "Always" or "Usually"
Modification from U.S. HWBI: Same as U.S. HWBI
129
-------
Appendix B: Variables in HWBI Processing
Table Bl. Description of HWBI processing output for Processing step 1: Correct data structure and temporal imputation
Variable
Values
Description
Calculation
Origin
METRICVAR
Character
Abbreviated name of HWBI Metric (80 total)
NA
Formatting
DOMAIN
Character
Name of HWBI domain (8 total)
NA
Formatting
INDICATOR
Character
Name of HWBI indicator (24 total)
NA
Formatting
POSNEGMETRIC
P, N
Describes whether or not a metric measures a positive or
P=if metric measures a positive component of
Formatting
negative component of well-being
well-being; N=if metric measures a negative
component of well-being. Negative metrics will
be corrected in processing step 4
FIPS
5-digit code
Unique, 5-digit Federal Information Processing Standards
NA
Formatting
(FIPS) county code
Year
2000 - 2013
Data value year
NA
Formatting
ORIG_FI PS
Numeric
Unique numeric code of finest spatial scale originally available
NA
Processing
code
by year (5-digit FIPS county code, 3- or 4-digit Metropolitan
step 1
Statistical Area (MSA) code, 7-digit Public Use Microdata Area
(PUMA) FIPS code, 2 digit state FIPS code, or 1-digit World
Values Survey (WVS) regional code). If 'measure' was carried-
forward (temporally imputed), then NA.
measure
Numeric
Metric value from processed input dataset or database for
Column name of original dataset or database;
Processing
finest available spatial scale, including carried-forward
measure will be subsequently modified during
step 1
(temporally imputed)
spatial imputation and to standardize units (e.g.,
percent and proportion) and spatial imputation
source
Character
Abbreviated name of data source
NA
Formatting
metricname
Character
Name of HWBI Metric (80 total)
NA
Formatting
Continued next page
130
-------
Table Bl. Continued.
Variable
Values
Description
Calculation
Origin
units
Character
Units of metric (dollars, index, minutes, percentage,
proportion, rate, years)
NA
Formatting
datafips
Numeric
Code representing the source of 'measure' (5-digit FIPS county If data are available, spatial scale assigned during
Processing
code
code, 3- or 4-digit Metropolitan Statistical Area (MSA) code, 7-
digit Public Use Microdata Area (PUMA) FIPS code, 2 digit
state FIPS code, or 1-digit World Values Survey (WVS) regional
code), including carried-forward (temporally imputed)
formatting. If temporally imputed, this is the
spatial scale used for imputation
step 1
datayear
2000-2013
Year of 'measure,' including carried-forward (temporally
If data are available, year assigned during
Processing
imputed)
formatting. If temporally imputed, this is the year
used for imputation
step 1
annualerror
Numeric
Estimate of annual imputation error (standard error of mean
For spatially unimputed: If year=data year, then
Processing
predicted value)
annualerror=0, if >2 observations then annual
error=standard error of the mean predicted value
(STDP) by metric and FIPS; otherwise
annualerror=NA
step 1
RUCC
Numeric
code
United States Department of Agriculture (USDA) Rural-Urban
Continuum Code (RUCC)
NA
Formatting
GINIJDXBAND
Numeric
code
U.S. Census GINI Index (GINI) for household income inequality
20% quantiles of U.S. GINI values 2000-2010
Formatting
131
-------
Table B2. Description of HWBI processing output for Processing step 2: Create metric fill data for spatial imputation
Variable
Values
Description
Calculation
Origin
FILL_SOURCE
Character
Geography used to summarize data. Data are summarized
(not subsetted) for a specific spatial scale (e.g., RUCC-GINI or
State) from all available data (finest spatial scale available,
which could be county-level, regional, state, etc.)
NA
Processing
step 2
METRICVAR
Character
Abbreviated name of HWBI Metric (80 total)
NA
Formatting
Year
2000-2013 Data value year
NA
Formatting
RUCC
Numeric
code
United States Department of Agriculture (USDA) Rural-Urban
Continuum Code (RUCC)
NA
Formatting
GINIJDXBAND
Numeric
code
U.S. Census GINI Index (GINI) for household income inequality
20% quantiles of U.S. GINI values 2000-2010
Formatting
FILL_MED
numeric
Median value of all measures by metric and year within
specified spatial scale
Median
Processing
step 2
FILL_MAD
numeric
Median absolute difference about the median by metric and
year within specified spatial scale
Median absolute difference about the median (In
SAS, this is the proc means MAD output; In R use
function mad(constant=l))
Processing
step 2
FILL_QSTD
numeric
Interquartile range standard deviation by metric and year
within specified spatial scale
Interquartile range standard deviation
Processing
step 2
FILL_MEAN
numeric
Mean value of all measures by metric and year within
specified spatial scale
Mean
Processing
step 2
FILL_ERROR
numeric
Standard Error of the mean of all measures by metric and year Standard error of the mean
within specified spatial scale
Processing
step 2
FILL_OBS
numeric
Number of observations by metric and year within specified
spatial scale
NA
Processing
step 2
132
-------
Table B3. Description of HWBI processing output for Processing step 3: Complete metric data with spatial imputations
Variable
Values
Description
Calculation
Origin
unimputed Y, N Describes whether or not a metric value is unimputed across
spatial scales. Data that have been "carried forward"
(temporally imputed) can still be unimputed as long as
original data is at the county-level
METRICVAR Character Abbreviated name of HWBI Metric (80 total)
DOMAIN Character Name of HWBI domain (8 total)
INDICATOR
Character Name of HWBI indicator (24 total)
POSNEGMETRIC P, N
Describes whether or not a metric measures a positive or
negative component of well-being
FIPS
Year
measure
5-digit code Unique, 5-digit Federal Information Processing Standards
(FIPS) county code
2000-2013 Data value year
Numeric If datafips is a county, metric value from processed input
dataset or database, including carried-forward (temporally
imputed). If datafips is regional or state-level, then measure
is spatial imputation
Y=if ORIG MEASURE is available at the county- Processing
level; N=if ORIG MEASURE is not available at the step 3
county level. Negative metrics will be corrected in
processing step 4
NA Formatting
NA Formatting
NA Formatting
P=if metric measures a positive component of Formatting
well-being; N=if metric measures a negative
component of well-being
NA Formatting
NA Formatting
Column name of original dataset or database and Processing
included temporal imputations from CORRECTED step 3
step. In processing step 3, spatial imputation
applied where datafips is not at county-level.
Measure will be subsequently modified to
standardize units (e.g., percent and proportion)
and spatial imputation
metricname Character Name of HWBI Metric (80 total)
NA
Formatting
Continued next page.
133
-------
Table B3. Continued
Variable
Values
Description
Calculation
Origin
units
Character Units of metric (dollars, index, minutes, percentage,
proportion, rate, years)
NA
Formatting
datafips
Numeric Code representing the source of 'measure' (5-digit FIPS county If data are available, spatial scale assigned during Processing
code code, 3- or 4-digit Metropolitan Statistical Area (MSA) code, 7- formatting. If temporally imputed, this is the step 1
digit Public Use Microdata Area (PUMA) FIPS code, 2 digit spatial scale used for imputation
state FIPS code, or 1-digit World Values Survey (WVS) regional
code), including carried-forward (temporally imputed)
datayear
2000-2013 Year of'measure,' including carried-forward (temporally
imputed)
If data are available, year assigned during Processing
formatting. If temporally imputed, this is the year step 1
used for imputation
annualerror Numeric Estimate of annual imputation error (standard error of mean
predicted value)
For spatially unimputed: If year=data year, then Processing
annualerror=0, if >2 observations then annual step 1
error=standard error of the mean predicted value
(STDP) by metric and FIPS; otherwise
annualerror=NA
RUCC
Numeric United States Department of Agriculture (USDA) Rural-Urban NA
code Continuum Code (RUCC)
GINIJDXBAND Numeric U.S. Census GINI Index (GINI) for household income inequality 20% quantiles of U.S. GINI values 2000-2010
code
Formatting
Formatting
MSA REGION
WVS REGION
Character U.S. Office of Management and Budget (OMB) Metropolitan NA
Statistical Area (MSA)
Character 2001 World Values Survey (WVS) region
NA
Formatting
Formatting
Continued next page.
134
-------
Table B3. Continued
Variable
Values
Description
Calculation
Origin
STATEABBR
PUMA REGION
Character U.S. Postal Service state abbreviation
Numeric
code
U.S. Census Bureau Public Use Microdata Area (PUMA)
NA
NA
Formatting
Formatting
FILL SOURCE
FILL MED
FILL MAD
FILL OBS
Character Geography used to summarize data. Data are summarized NA
(not subsetted) for a specific spatial scale (e.g., RUCC-GINI or
State) from all available data (finest spatial scale available,
which could be county-level, regional, state, etc.)
numeric Median value of all measures by metric and year within Median
specified spatial scale
numeric Median absolute difference about the median by metric and Median absolute difference about the median (In
numeric
ORIG MEASURE Numeric
year within specified spatial scale
Number of observations by metric and year within specified
spatial scale
SAS, this is the proc means MAD output; In R use
function mad(data, constant=l))
Sample size
Original metric value of data from processed input dataset or Measure' before spatial imputation
database, including carried-forward (temporally imputed)
datascale
Character Source of original data (finest available spatial scale)
Spatial level of datafips
Processing
step 2
Processing
step 2
Processing
step 2
Processing
step 2
Processing
step 3
Processing
step 3
135
-------
Table B4. Description of HWBI processing output for Processing step 4: Normalize and standardize metric values
Variable
Values
Description
Calculation
Origin
unimputed Y, N Describes whether or not a metric value is unimputed across
spatial scales. Data that have been "carried forward"
(temporally imputed) can still be unimputed as long as
original data is at the county-level
METRICVAR Character Abbreviated name of HWBI Metric (80 total)
DOMAIN Character Name of HWBI domain (8 total)
INDICATOR
Character Name of HWBI indicator (24 total)
POSNEGMETRIC P, N
Describes whether or not a metric measures a positive or
negative component of well-being
Y=if ORIG MEASURE is available at the county-
level; N=if ORIG MEASURE is not available at the
county level
NA
NA
NA
P=if metric measures a positive component of
well-being; N=if metric measures a negative
component of well-being. Negative metrics will
be corrected in processing step 4
Processing
step 3
Formatting
Formatting
Formatting
Formatting
FIPS
Year
measure
5-digit code Unique, 5-digit Federal Information Processing Standards
(FIPS) county code
2000-2013 Data value year
Numeric If datafips is a county, metric value from processed input
dataset or database, including carried-forward (temporally
imputed). If datafips is regional or state-level, then measure
is spatial imputation
NA
NA
Column name of original dataset or database and
included temporal imputations from CORRECTED
step. In COMPLETE step, spatial imputation
applied where datafips is not at county-level.
Measure will be subsequently modified to
standardize units (e.g., percent and proportion)
and spatial imputation
Formatting
Formatting
Processing
step 3
Continued next page
136
-------
Table B4. Continued.
Variable
Values
Description
Calculation
Origin
metricname
units
datafips
Character Name of HWBI Metric (80 total)
datayear
annualerror
Character
Numeric
code
All standardized metrics have units "Proportion" (scaled 0.1 ¦
0.9)
NA
Units have been modified from formatting
Formatting
Processing
step 4
Code representing the source of 'measure' (5-digit FIPS county If data are available, spatial scale assigned during Processing
code, 3- or 4-digit Metropolitan Statistical Area (MSA) code, 7- formatting. If temporally imputed, this is the step 1
digit Public Use Microdata Area (PUMA) FIPS code, 2 digit
state FIPS code, or 1-digit World Values Survey (WVS) regional
code), including carried-forward (temporally imputed)
2000 - 2013 Year of 'measure,' including carried-forward (temporally
imputed)
Numeric Estimate of annual imputation error (standard error of mean
predicted value)
spatial scale used for imputation
If data are available, year assigned during Processing
formatting. If temporally imputed, this is the year step 1
used for imputation
For spatially unimputed: If year=data year, then Processing
annualerror=0, if >2 observations then annual step 1
error=standard error of the mean predicted value
(STDP) by metric and FIPS; otherwise
annualerror=NA
RUCC
Numeric United States Department of Agriculture (USDA) Rural-Urban NA
code Continuum Code (RUCC)
GINIJDXBAND Numeric U.S. Census GINI Index (GINI) for household income inequality 20% quantiles of U.S. GINI values 2000-2010
code
Formatting
Formatting
MSA REGION Character U.S. Office of Management and Budget (OMB) Metropolitan
Statistical Area (MSA)
NA
Formatting
Continued next page
137
-------
Table B4. Continued.
Variable
Values
Description
Calculation
Origin
WVSREGION Character 2001 World Values Survey (WVS) region NA
STATEABBR Character U.S. Postal Service state abbreviation NA
PUMAREGION Numeric U.S. Census Bureau Public Use Microdata Area (PUMA) NA
code
Formatting
Formatting
Formatting
FILL SOURCE
FILL MED
FILL MAD
FILL OBS
Character Geography used to summarize data. Data are summarized NA
(not subsetted) for a specific spatial scale (e.g., RUCC-GINI or
State) from all available data (finest spatial scale available,
which could be county-level, regional, state, etc.)
numeric Median value of all measures by metric and year within Median
specified spatial scale
numeric Median absolute difference about the median by metric and Median absolute difference about the median (In
numeric
ORIG MEASURE Numeric
datascale
year within specified spatial scale
Number of observations by metric and year within specified
spatial scale
SAS, this is the proc means MAD output; In R use
function mad(constant=l))
Sample size
Original metric value of data from processed input dataset or Measure' before spatial imputation
database, including carried-forward (temporally imputed)
Character Source of original data (finest available spatial scale)
Spatial level of datafips
Processing
step 2
Processing
step 2
Processing
step 2
Processing
step 2
Processing
step 3
Processing
step 3
Continued next page
138
-------
Table B4. Continued.
Variable
Values Description
Calculation
Origin
outlier
0,1
Describes whether or not a metric value is an outlier (0=not Outlier if measure q3+(3*qrange), otherwise not outlier. Prior to step 4
normalization, we identified outlying values
falling beyond the far fences of a box-and-whisker
plot (i.e. less/greater than three interquartile
ranges from the 1st or 3rd quartiles)
MINVAL
numeric Minimum value per metric where outlier=0
Three interquartile ranges greater than the 3rd Processing
quartile; "Floor" step 4
MAXVAL
numeric Maximum value per metric where outlier=0
Three interquartile ranges less than the 1st Processing
quartile; "Ceiling" step 4
numeric Number of values per metric where outlier=0
Sample size
Processing
step 4
nALL
numeric Number of total values per metric
Sample size
Processing
step 4
scalefactor numeric How much of original range is being used after outliers are
removed (0=none, l=all)
|maxval-minval| / | maxvalall-minvalall |
Processing
step 4
Continued next page
139
-------
Table B4. Continued.
Variable
Values
Description
Calculation
Origin
impactfactor
numeric How many data points are affected by ceiling/floor
(nALL-n)/nALL (O=none, l=al
Processing
step 4
NUMERATOR
numeric
Numerator for standardization
measure-minval
Processing
step 4
DENOMINATOR numeric
Denominator for standardization
maxval-minval
Processing
step 4
METRIC VAL
numeric Standardized value of metric measure, corrected for positive If metric is positive then
Processing
(0.1-0.9) and negative metrics (metrics that contribute positively to
well-being are closer to 0.9)
METRIC_VAL=numerator/denominator; if metric step 4
is negative then METRIC_VAL=1-
(numerator/denominator) (Outliers < ql-
(3*qrange)=0.1; outliers (>q3+(3*qrange)=0.9;
corrected for Positive/Negative metric)
UNITS
Character All standardized metrics have units "Proportion" (scaled 0.1 ¦
0.9)
Units have been modified from formatting
Processing
step 4
140
-------
Office of Research and Development v>EPA
Washington, DC 20460 United States
Environmental Protection
Agency
EPA/600/R-16/363
------- |