United States Air and Radiation EPA420-P-99-007
Environmental Protection March 1999
Agency M6.IM.001
vvEPA MOBILES Inspection/
Maintenance Benefits
Methodology for
1981 through 1993 Model
Year Light Vehicles
DRAFT
> Printed on Recycled Paper
-------
EPA420-P-99-007
March 1999
for
1981
M6.IM.001
DRAFT
Edward L, Glover
Dave Brzezinski
Assessment and Modeling Division
Office of Mobile Sources
U.S. Environmental Protection Agency
NOTICE
This technical report does not necessarily represent final EPA decisions or positions.
It is intended, to present technical analysis of issues using data which are currently available.
The purpose in the release of such reports is to facilitate the exchange of
technical information and to inform the public of technical developments which
may form the basis for a final EPA decision, position, or regulatory action.
-------
EPA420-P-99-007
- Draft -
MOBILE6 Inspection / Maintenance Benefits Methodology
for 1981 through 1993 Model Year Light Vehicles
Report Number M6.IM.001
Last Revised February 22, 1999
Edward L. Glover
Dave Brzezinski
U.S. EPA Assessment and Modeling Division
NOTICE
This technical report does not necessarily represent final EPA decisions or positions. It is intended to present
technical analysis of issues using data which are currently available. The purpose in the release of such
reports is to facilitate the exchange of technical information and to inform the public of technical
developments which may form the basis for a final EPA decision, position, or regulatory action.
M6IM001.WPD DRAFT 1 Mar 24, 1999
-------
DRAFT
EXECUTIVE SUMMARY
List of Issues, Key Points, Assumptions and User Inputs
Regarding MOBILE6 I/M Credits
The methodology described in this document (M6.IM.001) covers 1981-93 model
year cars and light-duty trucks, and 1994 and 1995 model year cars and trucks which were
not certified to Tierl or later standards. It calculates separate I/M credits for running and
start emissions. I/M credits are based on a simple distribution model in which every
vehicle in the fleet is either a high emitter (FTP emission greater than 2 times HC or NOx
standards or 3 times CO standards) or a normal emitter. The emission levels of the high
and normal emitters are based on FTP data collected independently by EPA, AAMA and
API as part of the organizations' in-use vehicle emission assessment programs. The
frequency and distribution of high and normal emitters in the fleet is based on a large
database of EVI240 data collected in Dayton, Ohio in 1996 and 1997. The basic emission
levels used in the model are a function of vehicle mileage, vehicle technology, and model
year.
The basic assumption behind I/M is that a fraction of the high emitters in the fleet
are identified and repaired down to lower emission levels during the I/M process. This
process reduces the average emission level of the fleet. It is modeled using a mathematical
model which resembles a 'sawtooth'. The bottom of the "teeth" are the after repair
emission levels immediately following I/M, and the top of the "teeth" are the levels to
which the fleet deteriorates after one periodic inspection cycle, or a six month RSD /
change of ownership cycle.
MOBILE6 will allow various I/M scenarios to be modeled. Some of these are new
to the MOBILE model series. The others have all been changed or revamped in a
significant manner. MOBILE6 will allow for some new features.
New Features:
1. Internal operation - No external I/M credit files to attach to the main program for
1981 and later model year vehicles.
2. I/M credits given for the EVI240 test, the ASM tests, the Idle tests and OBD testing.
3. Custom user supplied cutpoints for EVI240 can now be entered directly in the
program. For example, the combination (1.5 g/mi HC, 55 g/mi CO, and 3.2 g/mi
NOx) can be entered for an EVI240 scenario.
M6IM001.WPD DRAFT 2 Mar 24, 1999
-------
DRAFT
4. Annual, Biennial, Triennial, and Change of Ownership I/M testing frequency can
now be modeled.
5. Ability to model up to five different I/M programs simultaneously.
6. Remote Sensing of High emitters can now be modeled.
7. Ability to model the exemption of the first "n" model years / ages in an I/M
program. The "n" can be up to the first 20 model years / ages.
8. User input and default values for non-compliance with testing requirements, and
cost waivers on failures can be specified.
9. I/M credits given for cost waivered vehicles.
10. Ability to model RSD Clean Screening and High Emitter Profiling exemptions from
an I/M program.
Development of Important Parameters
1. The I/M methodology and associated parameters presented in this document are
heavily based on two other EPA documents. These are "Determination of Running
Emissions as a Function of Mileage for 1981-93 Model Year LDV and LDT
Vehicles" - M6.EXH.001, and "Determination of Start Emissions as a Function of
Mileage and Soak Time for 1981-93 Model Year Light Duty Vehicles." -
M6.STE.003. The reader is encouraged to obtain these documents from the EPA
Web site and review them.
2. Grouping Parameters - Most of the grouping of the data was done by model year
and technology groups. Ported fuel injection (PFI) technology was split from
throttle body injection (TBI) and carbureted technology. Model year groups were
chosen based on engineering judgement regarding technology changes, or were
grouped based on similar certification emission standards.
3. Basic emission rate and I/M analyses were done for both cars and light trucks
separately. The same analysis approach was used for each vehicle type; however,
different model year grouping were selected for cars and trucks because of the
different certification standards which were in effect.
M6IM001.WPD DRAFT 3 Mar 24, 1999
-------
DRAFT
4. Basic Emission Rates - FTP emission factor data comes from significant EPA and
industry testing (3,000+ FTP tests), and was corrected for recruitment bias based
on IM240 testing from Dayton, Ohio (211,000 IM240 tests).
5. Average emissions of Normals and Highs for start and running emissions - FTP
data were used.
6. Identification Rate of High emitters - These are based on a sizeable database (900
vehicles) which received both the FTP and EVI240 tests at an EPA contractor
facility.
7. After I/M Repair Effects for running emissions - These are based on thousands of
EVI240 tests from Arizona on vehicles which were repaired to pass I/M.
8. After I/M Repair Effect for start emissions - These are based on FTP data collected
by EPA.
9. Sawtooth Methodology - It is from MOBILES. It assumes that vehicles repaired
as part of the I/M process deteriorate at the same rate as a fleet which does not have
an I/M program. However, unlike previous MOBILE models, the deterioration
varies over the entire mileage range of 0 to 300,000 miles.
10. Waiver Repair Levels - In MOBILE6, cost waivered I/M failures will get some
repair benefit. A value of a 20 percent reduction has been chosen. This value may
change between draft and final versions, if real data provides another value.
Stakeholders are encouraged to comment on this assumption, and provide any data
or rational for an alternative default value.
11. High Emitter Non-Compliance Rate - Set to a default value of 15 percent.
MOBILE6 will offer users the ability to enter alternative values. This is a generous
default which is based on extensive analysis of Arizona and Ohio I/M vehicles.
The analysis suggested higher rates (> 20 percent). It also includes high emitters
which do not show up for the initial I/M test. The fact that 15 percent has been
selected for use in the absence of user input does not constitute a policy by EPA to
allow the use of this value for SIP purposes. EPA will propose a policy on this
issue separately from this document. Stakeholders are encouraged to comment on
this assumption, and provide any data or rational for an alternative default value.
12. High Emitter Waiver Rate - Selected to be 15 percent of failures, and loosely
based on analysis of Arizona and Ohio I/M vehicles. The user will also have the
ability to enter an alternative value into MOBILE6. Also, the fact that 15 percent
M6IM001.WPD DRAFT 4 Mar 24, 1999
-------
DRAFT
has been selected for use in the absence of user input does not constitute a policy
by EPA to allow the use of this value for SIP purposes. EPA will propose a policy
on this issue separately from this document. Stakeholders are encouraged to
comment on this assumption, and provide any data or rational for an alternative
default value.
13. Remote Sensing Parameters - These are based on two reports published by EPA.
One report was on RSD identification of high emitters and the other was on RSD
clean screening effectiveness. RSD and Change of Ownership modeling is new to
the MOBILE model series, and requires several new inputs. However, its impact
is relatively minimal on the overall I/M credits or basic emission level rates.
14. Assume on average for the fleet that one RSD inspection to identify high emitters
is done per year on each vehicle in the I/M program. Field experience with RSD
suggests that this is an ambitious goal, and may require many vehicles to get dozens
of RSD tests per year; however, programs which manage to test more frequently
than this rate will not get additional credit. The user will also be allowed to enter
a specific RSD fleet coverage fractions for RSD high emitter identification and
RSD clean screening. The range of these fractions will be from 0 percent to 100
percent.
15. The default RSD or High Emitter Profile clean screening loss of credit is five
percent. However, the user of MOBILE6 is strongly encouraged to develop their
own estimate and use it as an input to the model. Stakeholders are encouraged to
comment on this assumption, and provide any data or rational for an alternative
default value.
16. Change of Ownership - Data from Wisconsin suggests that roughly 16 percent of
the testing annually is change of ownership testing. This translates into 8 percent
every six months, and is built into the change of ownership "sawtooth" algorithm.
17. MOBILE6 will assume that the ASM tests will have the same relative performance
to the EVI240 that they did in MOBILES. This is necessary because no new ASM
I/M test data matched with FTP data are available since MOBILES was released.
New Idle and 2500RPM/Idle test data are available and new performance estimates
have been computed, and will be installed in the MOBILE6 model. The ASM and
Idle I/M test performance in comparison to the EVI240 will be computed in the
MOBILE6 model by adjusting the I/M test identification rate (IDR) factors.
M6IM001.WPD DRAFT 5 Mar 24, 1999
-------
DRAFT
18. Both the ASM and Idle tests assume the same after I/M repair emission levels as the
IM240 tests. Only the IDR rates are changed. This assumption is currently under
review for the Idle and Idle/2500 RPM tests. The most likely change will be to
adopt the MOBILES repair effects for Idle tests rather than assume the Idle test has
the same repair reduction as the EVI240 test. The ASM test will continue to use the
same repair effect as the EVI240 test.
General Statement
This document and the important parameters mentioned in it are currently in
DRAFT status, and will likely remain in that status until mid-1999. This document will
also likely receive some revision following peer review and stakeholder review. As a
result, the I/M model, the basic emission rates, and the underlying parameters are all subj ect
to possible future revision. Comments regarding the modeling approach, important
parameters and assumptions are encouraged from stakeholders and other interested parties.
M6IM001.WPD DRAFT 6 Mar 24, 1999
-------
DRAFT
1.0 INTRODUCTION
This document describes EPA's new methodology for estimating exhaust emission
Inspection / Maintenance (I/M) credits. This includes the methodology for various tests
such as the IM240, the Idle test, the 2500 RPM/Idle test, and the ASM test. It includes the
methodology used for all cars and light trucks for model years 1981 through 1993, and for
non-Tier 1 cars and trucks for model years 1994 and 1995. The I/M credit methodology for
the pre-1996 model year will also be used for 1996 and later model years which receive
only exhaust I/M tests. This document also describes how credits and debits for the remote
sensing device (RSD) testing will be incorporated into MOBILE6. The I/M credits for the
pre-1981 model years are not being revised for MOBILE6. The I/M credits for post-1995
model years with OBD systems, and the evaporative emission I/M test credits will be
discussed in a separate documents "Determination of Emissions, OBD, and I/M Effects for
Tierl, TLEV, LEV, and ULEV Vehicles" - EPA document M6.EXH.007, and "Inspection
/ Maintenance Credits for Evaporative Control System Tests" - EPA document
M6.IM.003.
MOBILE6 will handle I/M credits differently than previous MOBILE models. One
major difference is the discontinuation of the TECHS model. The TECHS model was a
complex external FORTRAN program which calculated and exported the exact I/M credit
values. These credit values were then built into the MOBILES block data code or read as
an external file. The new credit methodology will instead be built into the MOBILE6 code,
and will operate automatically every time an I/M program is called by the MOBILE6
program. This change will give the MOBILE6 user the ability to vary the effect of
cutpoints and other program parameters through changes to the MOBILE6 input file. No
longer will it be necessary to develop special I/M credits using the TECHS model, and
attach them to the MOBILE program.
The new I/M credit methodology will also be updated to reflect the new basic
emission rates (see "Determination of Running Emissions as a Function of Mileage for
1981-1993 Model Year Light-Duty Vehicles-Report Number M6.EXH.001"). In addition
to being lower in magnitude, the new emission rates separate start and running emissions.
MOBILE6 will account for these emissions separately, and produce separate start and
running I/M credits.
This document is structured into five primary sections, and an Appendix section.
Section 2 briefly describes the databases used in the analysis and development of the
credits. Section 3 describes the methodology for development of the running exhaust I/M
credits based on the EVI240 test. Section 4 describes the periodic I/M methodology -
"sawtooth methodology", Section 5 describes the methodology for development of the start
exhaust I/M credits based on the EVI240 test. Section 6 describes the methodology for the
M6IM001.WPD DRAFT 7 Mar 24, 1999
-------
DRAFT
development of credits for the other types of I/M tests (Idle, 2500/Idle, and ASM). The
document also contains an extensive Appendix section which is listed A through H.
Appendices A and C contain illustrative examples of the modeling approach. Appendix
C contains some sample calculations. Appendix D contains the programmer's explanation
and adoption for coding purposes of the algorithm described in this document. Appendices
E through H contain statistical diagnostics for many of the parameters used in this model.
2.0 DATA
Four databases were utilized to develop the EVI240 based credits. The first database
was a large emission factor database which contained over 5,000 initial FTP tests on 1981
through 1993 model year cars. It was used in the I/M credit analysis to determine the
average emissions of the "Normal" emitting vehicles and the "High" emitting vehicles.
This is the same database which was used in generating the basic emission rates. It is
described in greater detail in "Determination of Running Emissions as a Function of
Mileage for 1981-1993 Model Year Light-Duty Vehicles" - report number M6.EXH.001.
The second database was a smaller I/M database. It was used to determine the high
emitter identification rates for the EVI240 test. It contained 910, 1981 and later cars and
trucks which had both an EVI240 test and a running LA4 test (derived from the FTP test).
It contained data from EPA emission factor testing in Ann Arbor, Indiana and Arizona in
which vehicles were randomly recruited and tested on both the FTP test and the EVI240 test.
This second vehicle emission database contains many of the same FTP / lane EVI240
test pairs that were used for the MOBILES I/M credits. In an attempt to update the
MOBILE6 credits with newer model year data, additional vehicle data with FTP / lab
EVI240 test pairs were added where FTP / lane EVI240 were not available. Use of a lab
EVI240 versus a lane EVI240 for I/M credit purposes introduces some additional uncertainty
in the analysis since a lab EVI240 test is less similar to an actual state conducted EVI240 I/M
test than a lane EVI240. However, inclusion of the FTP / lab test data, enabled the analysis
to include some post 1991 model year vehicles and additional light trucks rather than
extrapolate these points. Thus, EPA concluded that these benefits outweighed the slight
increase in uncertainty caused by using lab EVI240 data.
The third database was the Arizona EVI240 database from official state testing. It
contained several thousand before-and after-repair EVI240 tests, and was used to determine
the repair effects for the running LA4 EVI240 credits. It contains data from a special test
program that the State of Arizona conducts on a continuous basis to evaluate the
performance of their I/M program. In this program, vehicles are randomly selected to
M6IM001.WPD DRAFT 8 Mar 24, 1999
-------
DRAFT
receive the full IM240 test both initially, and if they fail, after all subsequent repair cycles.
The fourth database of about 970 EPA tested vehicles contained both EVI240 and
FTP data before and after repair. It was used to calculate the effects of repair on start
emissions. It is documented in EPA document M6.EVI.002.
The RSD credits and coverage parameters described in this document are based on
extensive RSD testing at many locations. Details regarding the RSD data, and the analysis
performed to determine the RSD credits can be found in EPA documents: "User Guide and
Description for Interim Remote Sensing Program Credit Utility - EPA420-R-96-004", and
"Draft Description and Documentation for Interim Vehicle Clean Screening Credit Utility -
EPA420-P-98-008".
3.0 I/M ALGORITHM FOR RUNNING EMISSIONS
3.1 Definition of Categories
The basic purpose of I/M is to identify and repair high emitting vehicles with broken
emission control systems. These types of vehicles are termed "high" emitters, and typically
have average emission levels which are considerably higher than the overall mean emission
levels. The remainder of the fleet is considered to be the "normal" emitters. These are low
and average emitting vehicles, and their emission control systems are generally functioning
properly. The overall fleet emission factor is assumed to be a weighted average of the high
and normal emitters. For comparison, the use of two emitter classes differs from the
methodology used in the previous TECHS and MOBILES models. In those models, there
were four emitter classifications (Normal, High, Very High, and Super).
The MOBILE6 model will generate specific I/M credits based on pollutant, model
year group, and technology type. Credits for the three pollutants HC, CO, and NOx will be
produced. Also, credits for the 1981 through 1993 model years will be stratified into seven
separate groups. These are: 1988-93 (PFI), 1988-93 (TBI), 1983-87 (FI), 1986-89
(CARB), 1983-85 (CARB), 1981-82 (FI), and 1981-82 (CARB). PFI means ported fuel
injection, TBI means throttle body fuel injection, (FI) means all closed-loop fuel injected,
and (CARB) means closed-loop carbureted and all open-loop vehicles combined together.
3.2 General I/M Algorithm
Figure 1 is a general graphical view of the I/M algorithm for running emissions.
Specific algorithms for each of the model year / technology / pollutant groups will be
M6IM001.WPD DRAFT 9 Mar 24, 1999
-------
DRAFT
programed into the MOBILE6 model. Four lines are shown in Figure 1 which show the
basic emission rate, the normal emitter emission rate, the high emitter emission level, and
the after repair emission levels of the high emitters which were identified and repaired. The
basic emission rate is shown as Line A. This line represents the average emissions of the
fleet without an I/M test. It includes both the normal vehicles and the high emitting
vehicles.
Line B in Figure 1 represents the average emissions of the normal vehicles. These
are the vehicles which are very unlikely to fail any EVI240 test cutpoint in the range used
by I/M programs, and should not require any significant emission related repair if they did
fail. The line is shown as a linear function of mileage to reflect the gradual deterioration
that normal vehicles experience due to general wear. In the data analysis these vehicles
were defined as normal emitters for a specific pollutant if their FTP HC emissions were less
than twice the applicable new car certification standard, or their FTP CO emissions were
less than three times the applicable new car certification standard, or their FTP NOx
emissions were less than twice times the new car certification standard. In MOBILE6, it
is assumed that these vehicles never fail I/M; no repair adjustment are made to them.
Line C in Figure 1 represents the average emissions of the high vehicles. These are
the vehicles which likely have "broken" emission control systems, and that should fail the
EVI240 test cutpoint, and receive repair. In the data analysis these vehicles were defined as
high emitters for a specific pollutant if their FTP HC emissions or FTP CO emissions
exceeded twice or three times the applicable new car certification standard, respectively,
or their FTP NOx emissions were two times the new car certification standard. Because
high NOx emissions often occur with low HC and/or low CO emissions, and sometimes
even HC can be high and CO normal, the three categories were kept separate. Thus, a
vehicle could be a high HC emitter, but a normal CO and NOx emitter.
The selection of twice or thrice FTP certification standards for the boundary level
between normals and highs is an engineering choice based on the literature on I/M and
repair. Other reasonable boundary levels could also have been chosen. No formal analysis
was done to prove that these levels were optimum. One of the reasons they were chosen
is because they were used in MOBILES, and have generally been shown in the past to be
a good dividing point between high emitting broken vehicles and lower emitting vehicles
which are not broken. Simple statistical analysis done on the data indicate that the two
means are statistically different.
Line D represents the average emissions of the portion of high emitting vehicles that
are identified and repaired because of the I/M process. This line is calculated as a function
of vehicle age, and is a percentage (e.g., 150%) of Line B. The portion of the fleet which
is identified by I/M will be repaired to a lower level on average. However, this level is not
M6IMOO1.WPD DRAFT 10 Mar 24, 1999
-------
DRAFT
as low on average as the average of the normal vehicles. The justification for this
assumption was an analysis of Arizona EVI240 before and after repair data collected during
1995 and 1996. (See EPA report EPA-420-R-97-001 "Analysis of the Arizona EVI240 Test
Program and Comparison with the TECHS Model" for a description of this dataset).
3.3 Calculation of Basic Running LA4 Emission Rates
Line A in Figure 1 represents the basic non-I/M emission rate for a given
combination of vehicle type / pollutant / model year group / technology group. The units
represented in Figure 1 are running LA4 emissions in grams / mile. The calculation
FIGURE 1
GENERAL I/M CREDITS SCHEMATIC
LA4
EMISSIONS
------ D
MILEAGE
M6IM001.WPD DRAFT
11
Mar 24, 1999
-------
DRAFT
methodology and databases used to determine these emission rates are fully documented
in the report "Determination of Running Emissions as a Function of Mileage for 1981-1993
Model Year Light-Duty Vehicles," report M6.EXH.001. The reader is encouraged to
review this document for more details. Selected emission rates were taken from
M6.EXH.001 and used in this current report as examples.
3.4 Calculation of Running LA4 Emission Rates for Normal Emitters
Line B in Figure 1 represents the average emission rates for Normal emitters. These
are the low emitting vehicles in the fleet which should not fail an I/M program. Line B was
calculated by least squares regression of the emissions of the normal emitters versus
mileage in the FTP dataset. Sample sizes were satisfactory in all cases. The regression was
done for each pollutant / model year / technology group. The regression coefficients for
cars are shown in Table la and light trucks in Table Ib. The column labeled ZML contains
the zero mile coefficients, and the column DET contains the deterioration coefficients
(slope) from the regressions (units are grams per mile per IK miles). A sample scatterplot
of the car data and the regression line is shown in Figure A-l through A-3 in Appendix A.
Table la
Regression Coefficients for RUNNING LA4 Emissions from Normal Emitter Cars
MY
Group
1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group
PFI
TBI
FI
Carb
Carb
FI
Carb
HC Coefficients
ZML
0.0214
0.0042
0.0942
0.0774
0.1266
0.0970
0.1539
DET
0.001385
0.001701
0.001439
0.000812
0.001214
0.002250
0.001271
CO Coefficients
ZML
0.4588
0.0000
1.4448
0.5666
0.7276
1.5762
1.3932
DET
0.02293
0.01990
0.01959
0.01371
0.01691
0.02150
0.01389
NOx Coefficients
ZML
0.2006
0.2253
0.4798
0.4960
0.5555
0.4597
0.5834
DET
0.00376
0.00381
0.00188
0.00170
0.00273
0.00633
0.00233
M6IM001.WPD DRAFT
12
Mar 24, 1999
-------
DRAFT
Table Ib
Regression Coefficients for RUNNING LA4 Emissions from
Normal Emitter Light Trucks
MY
Group
1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group
PFI
TBI
FI
Carb
Carb
HC Coefficients
ZML
0.02989
0.04664
0.13384
0.26835
0.49182
DET
0.002376
0.002998
0.003280
0.002701
0.006485
CO Coefficients
ZML
0.4927
0.7663
1.6222
1.3553
7.4202
DET
0.02678
0.03442
0.04311
0.06660
0.03293
NOx Coefficients
ZML
0.3024
0.3150
0.3150
1.2872
1.6159
DET
0.003904
0.003171
0.003171
0.0001
0.000025
3.5 Calculation of Running LA4 Emission Rates for High Emitters
Line C in Figure 1 represents the average emission rates for High emitters. These
are the vehicles in the fleet which likely have problems with their emission control systems,
and have emission levels which are considerably higher than the vehicles which do not have
problems. In the analysis they were defined as those vehicles exceeding either twice FTP
standards for HC or three times FTP standards for CO or twice NOx standards. The line
is a flat horizontal line because the emissions of a high emitter is not likely to be a strong
function of mileage. Regressions of the high emitter emissions versus mileage were done.
However, the relatively small sample sizes of high emitters make regression determined
mileage coefficients unreliable indicators of actual behavior. Various analyzes of failing
cars in I/M programs also support the use of a flat emission rate for high emitters.
Instead, on many new vehicles, if something goes seriously wrong with the emission
control system that is likely to immediately lead to high emissions, it is likely to be fairly
random in occurrence (i.e., not mechanical wear in the carburetor that creates large numbers
of high emitters over time, or built-in obsolescence at a particular mileage). However, one
weakness of this simplified approach is that a certain percentage (extremely small) of the
brand new vehicles will be modeled as being high emitters. This result occurs because at
zero miles, the regression developed estimate of normal emitter's emission level is below
the FTP and Ohio data developed estimate of the corresponding mean fleet emission level.
M6IM001.WPD DRAFT
13
Mar 24, 1999
-------
DRAFT
Table 2a shows the average emissions of the high emitters (cars only) for the 21
pollutant / model year / tech groups. Trucks are shown in Table 2b. Because of the small
sample size of high emitters in most groups, some model year / technology groups were
combined into another model year group, and an overall mean was computed for the
combined group. For the cars and for each pollutant, the 1986-89 Carb and the 1983-85
Carb were combined and averaged together. Likewise the 1981-82 Carb and 1981-82 FI
Car groups were combined and the emissions from the high emitters were averaged
together. For the trucks, all of the fuel injected trucks were combined together and a
common mean high emitter emission level was computed for each pollutant. This
combination had the effect of producing more consistent means across groups. The high
emitter HC emission level for the 1988-93 MY PFI group is also a special case. For this
group it was thought that the average high emitter emission level was too low because it
caused the average high emitter level to be lower than the normal emitter level at fairly low
mileages. It was increased from 1.10 g/mi HC to 1.74 g/mi HC by adding one very high
emitting 1987 model year vehicle to the 1988-93 model year PFI group.
The impact of this approach of averaging between groups and adding selected
vehicles to particular groups is that some high emitting vehicles contribute to the average
high emitter level of their own model year group, and to another model year group. This
does not affect the non-I/M running emission estimates because the normal and high emitter
split is not used to calculate the non-I/M estimates. However, it does affect the I/M
emission rate and I/M benefits because it changes the portion of a particular model year
group's emission distribution between normals and highs. This changed emission
distribution will affect the fraction of fleet emissions in MOBILE6 which are identified and
repaired by I/M. It is difficult to predict the size of the emission impact because it
simultaneously increases the average high emitter average, but decreases the fraction of
high emitters in the fleet. This change will also impact the start emissions and the start I/M
credits because it changes the fraction of high start emitters in the fleet (fraction of start
high emitters is equal to the fraction of running LA4 high emitters), but does not affect the
average start high emitter level.
An analysis of the Ohio EVI240 data was also done to try and estimate the high
emitter levels for running LA4 and start emissions. This was done because of the small
numbers of high emitters in the EPA and AAMA FTP (running LA4 and Start) data
samples. In this analysis, a large sample of Ohio vehicles were segregated into normal and
high emitters, and the average high emitter emission levels were determined and compared
with the FTP based estimates. They compared favorably. However, the analysis was
plagued with uncertainties such as how to separate the normals from the highs when FTP
data are not available, the inability to split PFI from TBI in the Ohio EVI240 data, a
questionable transformation of EVI240 results into running LA4 and start emissions, and
unknown and possibly inconsistent conditions between lab testing and EVI240 lane testing.
M6IMOO1.WPD DRAFT 14 Mar 24, 1999
-------
DRAFT
Because of these problems the Ohio IM240 data were not used to estimate the average high
emitter emission levels.
Table 2a
Mean RUNNING Emissions of High Emitter Cars
MY Group
1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group
PFI
TBI
FI
Carb
Carb
FI
Carb
HC Mean
1.740
3.394
2.372
1.845
1.845
2.372
2.372
CO Mean
36.106
46.527
37.933
27.653
27.653
37.933
37.933
NOx Mean
2.846
2.872
2.951
2.872
2.872
2.951
2.951
Table 2b
Mean RUNNING Emissions of High Emitter Light Trucks
MY Group
1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group
PFI
TBI
FI
Carb
Carb
HC Mean
2.120
3.241
2.446
2.012
3.710
CO Mean
33.283
33.283
43.870
39.415
80.726
NOx Mean
2.846
2.846
2.846
4.988
5.014
M6IM001.WPD DRAFT
15
Mar 24, 1999
-------
DRAFT
3.6 Calculation of After Repair Percentages and Emission Levels
Line D in Figure 1 represents the average after repair emission level of high emitters
that are properly identified and repaired. In comparison, Line C represents those high
emitting vehicles that are not identified and repaired properly, or belong to owners who
evade the program after failing the initial test. Line D is calculated by scaling up the
normal emitter emission level (Line B) using a multiplicative factor process which is a
function of age, pollutant and cutpoint level. The normal emitter emission level is the basis
for the after repair emission level, and is the lowest emission level to which high emitting
vehicles can be repaired after adjustment for age and mileage. This assumes that the I/M
process on average does not turn aged vehicles into brand new ones. However, the process
will allow an I/M program to claim full credit for fixing vehicles with definitive problems
such as a bad oxygen sensor.
3.6.1 After I/M Repair Multiplicative Adjustment Factor
The after I/M repair multiplicative adjustment factor is a function of vehicle age and
I/M cutpoint. It is calculated using a two step process. The first step is to calculate the
multiplicative adjustment factor for the standard set of EVI240 cutpoints which the State of
Arizona used in its EVI240 program. These are the phase-in cutpoints of 1.2 g/mi HC / 20
g/mi CO and 3.0 g/mi NOx. The second step involves computing and applying another
ratio which is a function of EVI240 cutpoint. It will allow the MOBILE6 program to assign
a different after repair emission level as a function of EVI240 cutpoint. The combined after
I/M repair multiplicative adjustment factor is multiplied by the normal emitter emission
level to calculate the after repair emission levels.
Phase-in Cutpoints
Equations 1 through 3 are the multiplicative adjustment factors used to calculate the
after repair emission level for HC, CO and NOx under phase-in cutpoints. They were
calculated from a large sample of Arizona EVI240 data. The same coefficients are used for
both cars and light trucks. The percent after repair I/M emission levels for the high emitters
which were identified by I/M and repaired were developed by: (1) Stratifying the sample
by age into 15 groups (ages 1 through 15); (2) Computing for each age group the average
emission level of the vehicles passing their initial Arizona I/M test; (3) Computing for each
age group the after repair passing emission values of the Arizona I/M failures; (4)
Computing for each age group the ratio of the emissions of the repaired high emitters over
the emissions of the initial passing vehicles; (5) Regressing the ratios versus age for each
of the three pollutants to produce Equations 1 through 3.
M6IMOO1.WPD DRAFT 16 Mar 24, 1999
-------
DRAFT
Equations 1 through 3 are used to produce Line D for the phase-in outpoints
(1.2/20/3.0) by following the two steps.
First, Line D is calculated as a percentage of Line B using Equations 1 through 3.
HC ratio = 2.2400 - 0.07595 * (vehicle age) Eqn 1
CO ratio = 2.1582 - 0.07825 * (vehicle age) Eqn 2
NOx ratio = 1.6410 - 0.04348 * (vehicle age) Eqn 3
In these equations, vehicle age ranges between 1 and 15 years, and the percentage
value at 15 years is used for all ages greater than 15.
Second, the percentage values calculated in Eqns 1 through 3 (i percentage in Eqn
4) are transformed into emission units by multiplying the percentage values by the emission
values in Line B (average emission of the normal emitters) using Eqn 4. The emission level
of the Normals is a function of mileage.
After repair emissions pollutant i = i percentage * Emissions of Normals Eqn 4
Other Cutpoint Combinations
Equations 1 through 4 are used to produce the after repair emission levels for an
EVI240 program which uses the phase-in cutpoints of 1.2/20/3 for HC, CO, and NOx
respectively. Another adjustment factor is used to compute after repair emission levels for
other cutpoints. It is a multiplicative factor which proportionally increases or decreases the
after repair emission level computed for the 1.2/20/3 phase-in cutpoints to account for
tighter or looser cutpoints.
The factor used to compute the after repair emission level for cutpoints other than
1.2/20/3 phase-in cutpoints is based on a limited amount of vehicle repair data collected
by EPA in past testing programs. It was utilized to overcome the limitation of repair data
collected at only one set of cutpoints in Arizona. This dataset was the same one used to
develop MOBILES repair effects and technician training I/M credits. The repair effects
dataset which was used consists of 273 vehicles from model years 1981 through 1992
tested by an EPA contractor in South Bend, Indiana and at the EPA lab in Ann Arbor, MI.
All of these vehicles had before and after repair EVI240 and FTP tests. The sample of
vehicles were repaired to various FTP emission level targets. None of the after repair
results included a catalyst replacement.
M6IMOO1.WPD DRAFT 17 Mar 24, 1999
-------
DRAFT
The principal goal of the data analysis was to determine as a function of IM240
cutpoint, the FTP after repair emission levels of vehicles which initially failed the EVI240
tests and were repaired to pass the EVI240 test. For MOBILES, this analysis was done for
seven different HC/CO cutpoint combinations and for five NOx cutpoints. These
combinations are repeated in this document because they are the only after repair FTP data
for a variety of cutpoints which currently exists. These cutpoint combinations are shown
in Tables 2c and 2d. Also, shown in Tables 2c and 2d are the after repair emission levels
for each cutpoint combination group, and the ratio of a given after repair emission level to
the after repair emission level at 1.20 g/mi HC / 20 g/mi CO. For NOx, the individual
cutpoint groups are ratioed to the 3.0 g/mi NOx group.
M6IMOO1.WPD DRAFT 18 Mar 24, 1999
-------
DRAFT
Table 2c
FTP After Repair HC and CO Emission Levels and Ratios
versus EVI240 HC/CO Cutpoint Combination
HC Cutpt
(g/mi)
1.2
0.8
0.6
0.6
0.6
0.4
0.4
CO Cutpt
(g/mi)
20
15
15
12
10
10
15
After
Repair
HC (g/mi)
1.26
1
0.88
0.87
0.86
0.78
0.74
After
Repair
CO (g/mi)
13.46
11.85
11.94
11.15
10.50
11.30
11.71
HC Ratio
1.00
0.79
0.70
0.69
0.68
0.62
0.59
CO Ratio
1.00
0.88
0.89
0.83
0.78
0.84
0.87
Table 2d
FTP After Repair NOx Emission Levels and Ratios
Versus NOx EVI240 Cutpoint
NOx Cutpt
(g/mi)
1
1.5
2
2.5
3.0
After Repair
NOx (g/mi)
0.91
1.22
1.48
1.68
1.86
NOx Ratio
0.489
0.656
0.796
0.903
1.000
For MOBILE6, the ratios data in Tables 2c and 2d were regressed versus HC, CO
and NOx outpoint to produce an after repair emission level ratio for any HC, CO or NOx
cutpoint (within the range allowed by MOBILE6) which the user may enter in MOBILE6
(the MOBILE6 user is no longer restricted to a set of seven cutpoint combinations). A least
M6IM001.WPD DRAFT
19
Mar 24, 1999
-------
DRAFT
squares linear regression was used to produce the relationships for both HC/CO and NOx.
The regression coefficients are shown in Table 2e. The equation form for the HC Ratio
and the CO Ratio are:
Ratio =
For NOx it is:
Ratio =
A*HCCut + B*COCut
B * NOCut + C
Eqn3b
Eqn3c
A linear regression was used instead of some other functional form because it produced
high r-squared values (0.99 for HC and NOx and 0.95 for CO). Also, note that the highest
IM240 cutpoint for HC and CO are 1.2 and 20 g/mi. Repair effects at cutpoints higher than
these will be linear extrapolation.
Table 2e
Regression Coefficients for Repair Effects Ratios
Ratio
HC Ratio
CO Ratio
NOx Ratio
A
0.4990
0.0249
B
-l.Olle-04
0.0168
0.2538
C
0.398
0.620
0.2613
rA2
0.996
0.950
0.993
3.6.2 Application of the After Repair Adjustment Factors
The ratio equations are used in MOBILE6 to compute the after repair emission
levels for cutpoints which are different from the standard 1.2 / 20 / 2.0 cutpoints used by
Arizona. This is done by multiplying Equations 1 or 2 or 3 by Equation 3b or 3c to produce
the repair effects ratio for the non standard (1.2/20/2.0) cutpoint. The final repair level is
obtained by multiplying this ratio by the appropriate normal emitter emission level line
(Line B). The normal emitter emission level is used as the final after repair emission level
if it is larger than the calculated after repair emission.
The following example calculation of the after repair HC emission level for an
HC/CO cutpoint combination of 0.80g/mi HC and 15 g/mi CO is shown below for clarity.
M6IM001.WPD DRAFT
20
Mar 24, 1999
-------
DRAFT
Aft Repair HC = (2.24-0.07595*age) * (0.4990*0.8g/mi - 1.01e-04*15.0g/mi + 0.398) * Norm_ave
where
Norm_ave is the average emissions of the normal emitters. It is a function of mileage and
technology/model year group. For an eight year old 1990 PFI vehicle at 100,000 miles it
is: 0.0214 + 0.001385 * 100 = 0.159 g/mi Running HC.
0.8g/mi HC is the HC cutpoint; 15.0g/mi is the CO cutpoint.
Substituting the value of 0.159 g/mi and 8 years old into the After Repair HC
equation produces an after repair emission level ofO.206 g/mi running HC at a cutpoint of
0.80 g/mi HC and 15 g/mi CO for an eight year old vehicle with 100,000 miles. This
compares with an after repair emission level for the same age and mileage of 0.260 g/mi
running HC at a cutpoint combination of 1.2/20 g/mi HC/CO. In this example, the after
repair emission level (0.206 g/mi HC) is above the value of the normal emitter (0.159 g/mi
HC). However, if the calculation produced a value which was lower, then the normal
emitter value would be used.
3.6.3 Discussion of the After Repair Adjustment Factors
This approach attempts to utilize the large sample of before and after repair EVI240
data collected in Arizona. These data are an improvement over the MOBILES assumptions
since they are a large sample, and are representative of the actual I/M experience. The in-
use data reflects the fact that regular commercial mechanics performed the repairs under
actual cost conditions. Also, the repairs were targeted to passing the actual state EVI240
test. Many of these technicians also received some training and orientation to the EVI240
program provided or encouraged by the State of Arizona prior to its implementation. The
principal assumption underlying this approach is the ratio between the after repair EVI240
emission level and the emission level of the vehicles passing the state EVI240 test is the
same as the ratio of the after repair running LA4 emission level and the normal emitter
running LA4 emission level. This is not an unreasonable assumption; however, there are
potential differences between the unpreconditioned EVI240 and the preconditioned running
LA4 test.
One drawback to the approach is that the Arizona data (and other states' data) were
available at only one cutpoint level (phase-in cutpoints). This made it impossible to
determine the sensitivity of repair levels to the EVI240 cutpoint. To overcome this obstacle
the previous FTP databases used for MOBILES were used to make the after repair effects
a function of cutpoint. A drawback to the use of these FTP data is that they are a relatively
small sample, the repairs were often performed by expert emission control system
M6IMOO1.WPD DRAFT 21 Mar 24, 1999
-------
DRAFT
technicians rather than commercial technicians, cost was usually not a factor in the repairs,
and specified numerical repair targets based on the FTP test were used. Also, running LA4
were not available so the FTP data were used directly under the assumption that the ratio
between cutpoints is same for the FTP and the running LA4.
3.6.4 Technician Training Effects
MOBILES had built-in I/M credits available for EVI240 programs which conducted
some form of technician training for people involved in I/M repairs. In MOBILE6, the after
repair emission levels discussed previously in Section 3.6 already include the effects of
technician training. This is because Arizona conducted a technician training program prior
and during implementation of their EVI240 program from which the repair effects data are
based.
Thus, it is proposed for MOBILE6 that the default after repair emission levels are
those 'with technician training'. For I/M programs which do not conduct a technician's
training program - 'w/o technician training', the after I/M repair emission levels will be
increased by the percentages shown in Table 2f.
The percentages shown in Table 2f are based on a limited study done by EPA to
evaluate technician training in an EVI240 program. In the program, eleven experienced
technicians in Arizona were trained on the eve of the EVI240 implementation in 1995 to
repair emission failures using a training program developed by Aspire, Inc., and taught by
an expert emission control system technician/trainer under EPA contract. Each participant
received the training and three vehicles to repair following the training. Unfortunately,
budget limitations prevented a good pre-training baseline of the technicians' performance
to be established. The study is fully documented in SAE Paper 960091.
The emission results shown in columns 2 and 3 of Table 2f are EVI240 test results
in units of grams per mile. The Student Tech column shows two numbers. The first
number is the before any repair emission level. It is shown for comparison only, and to
demonstrate that the technicians made sizeable emission reductions from repairs. The
second number is the average after repair EVI240 emission levels of the vehicles after the
students completed their work. The Master Tech column shows the average after repair
EVI240 levels after the instructor completed any additional repairs which were needed to
bring the vehicle into complete compliance. On a few vehicles this included a new catalytic
converter.
The % Difference column is the percent difference between the after repair student
tech and the after repair master tech emission results with the after repair master tech results
M6IM001.WPD DRAFT 22 Mar 24, 1999
-------
DRAFT
as the basis. It demonstrates the potential difference in performance between a master tech
and a trainee (journeyman) tech. It is proposed for MOBILE6 to calculate the 'w/o tech
training' after repair levels (w/o means without) by increasing the 'with tech training' values
by the % Difference values in Table 2f.
Table 2f
Technician Training Emission Effects
Pollutant
HC
CO
NOx
Master Tech
IM240 (g/mi)
0.38
3.00
1.11
Student Tech
IM240 (g/mi)
2.16/0.68
26.4/8.21
3.66/1.54
% Difference
78%
174%
39%
Use of these limited data in MOBILE6 for technician training effects requires two
important assumptions. First, that the after repair levels developed in the previous sections
already contain the effects of technician training. This is a reasonable assumption since
Arizona did institute a technician training program, and the after repair emission levels are
at relatively low levels. Second, that the difference on a percentage basis between the
master tech performance and the student tech performance is the same as the percentage
difference between the with and w/o technician training in the overall fleet. This
assumption is a little tenuous since the performance of typical trained technician is not as
high as the master tech in this study. This would have a tendency to produce a larger
percentage increase than in actuality. On the other hand, the student tech results were
collected after the training rather than before the training, and do not strictly represent un-
trained technicians. This factor would have a tendency to produce a smaller percentage
increase than in actuality.
Overall, the two factors discussed above might tend to cancel each other out.
However, because of these problems, the MOBILE6 program will allow optional user input
of 'w/o technician training' emission increase percentages. These will be for users who do
not have or plan to have a technician training program as part of their I/M program, but can
nevertheless estimate the detrimental impact of not having one through engineering
judgement, use of data from other I/M programs or other methods.
M6IM001.WPD DRAFT
23
Mar 24, 1999
-------
DRAFT
3.7 Waiver Repair Line
Not shown in Figure 1 is the waiver vehicle repair line. However, this line falls
between the high emitter level and the after proper repairs line. These are failing vehicles
which received a waiver from program requirements because a minimum amount of money
was spent on unsuccessful or only partially successful repairs. Typically, in most I/M
programs this means that between $200 and $450 was spent on the vehicle, and it still fails
the I/M test. The waiver repair line is below the high emitter line, despite the vehicle's
failing status, because even some limited or ineffective repair translates into reduced
emissions on average.
Because no analysis has yet been conducted on data from operating EVI240
programs to estimate the after I/M emission level of vehicles which were waived from the
requirement to pass the test, an assumed reduction percentage will have to be used, or the
individual user will have to provide a value. The default value will be a 20 percent
reduction from the high emitter line for all pollutants. The user will also have the option
of providing their own value(s) separately for each pollutant based on program data. If
EPA completes such an analysis between draft and final versions of this portion of
MOBILE6, or receives one in the comment period, the default value may be changed to
another number.
3.8 Percentage of High and Normal Emitters in the Fleet
Figure 1 shows in a general sense the overall fleet average emission level, the
average emissions of the normal emitters, and the average emissions of the high emitters.
The fleet average emission level was developed independent of the I/M credits, and the
methodology for its development is documented in EPA document M6.EXH.OO 1. In-order
to compute the I/M credits, the percentage of high emitters and normal emitters in the fleet
must also be calculated. Fortunately, this is an easy task since the average emission rate
is a weighted average of the normal emission rate and the high emission rate. The
weighting factors are simply back calculated to make this true at all odometers.
The fraction of High and Normal emitters is calculated for each combination of
vehicle type / pollutant / model year / technology group using the following general
equations.
Where:
Highs = fraction of High emitters at each age point
Normals = fraction of Normal emitters at each age point
LA4 is the average emission rate at each age point (determined in M6.EXH.001)
M6IM001.WPD DRAFT 24 Mar 24, 1999
-------
DRAFT
High_ave is the high emitter emission average at each age point
Norm_ave is the normal emitter emission average at each age point
Highs + Normals = 1 Eqn 5
and
LA4 = High_ave * Highs + Norm_ave * Normals Eqn 6
Solving for the variables Highs and Normals produces:
Highs = (LA4 - Norm_ave) / (Highave - Norm_ave) Eqn 7
Normals = 1 - Highs Eqn 8
For the model year groups of 1981-82 and 1983-85 HC and CO emissions, it was
found that the base emission factors at higher mileage levels become higher than the
average emissions of the high emitters. It occurs because at high mileages the basic
emission factors are data extrapolations. However, under the structure of the model, this
is not possible, and it implies that the fleet contains more than 100 percent high emitters.
To overcome this inconsistency, it was assumed that the average base emission factors
could not continue to rise after it reaches the average of the high emitters, and that it would
be set to the average of the high emitters. Typically, the cross-over point is between
150,000 and 200,000 miles, and after this point is reached, it is assumed that the percentage
of highs in the fleet for this model year group / technology is 100 percent. This flattening
of the emission factor line at very high mileages is consistent with some remote sensing
studies. A physical explanation would be that while some surviving vehicles continue to
deteriorate, the worst emitters are progressively scrapped out of the fleet in the high mileage
range.
3.9 High Emitter Identification Rates
The high emitter identification rate (DDR) represents the ability of an I/M test to
identify (fail) vehicles which are high emitters. It is represented as the percentage of the
total sum of emissions from the high emitters in the fleet. For example, the DDR would be
100 percent if it identified all of the running LA4 emissions from the high emitters in the
fleet. For the HC and CO I/M credits, the IDR is a function of the EVI240 HC and CO
cutpoints. For NOx I/M credits, it is a function of the NOx cutpoints only. In MOBILE6,
the user will be able to supply the exact EVI240 cutpoints which are desired, and the
program will automatically calculate the IDR and the credits. The EVI240 cutpoints will
M6IM001.WPD DRAFT 25 Mar 24, 1999
-------
DRAFT
need to be in the ranges: HC: 0.50 to 5.0 grams/mile; CO: 5.0 to 100.0 grams/mile; and
NOx: 1.0 to 5.0 grams/mile.
The I/M IDRs equations were calculated from the 910 vehicle database that
contained vehicle emission data from both running LA4 tests (FTP tests) and EVI240 tests
on lane fuel on cars and trucks. Cars and trucks will have the same DDR rates in MOBILE6
at a given cutpoint. However, separate cutpoints will be allowed for cars and trucks and
for each model year in a given MOBILE6 run. The analysis to develop the IDRs consisted
of several steps:
(1) The sample was split into two groups - the high HC and CO emitters, and the
high NOx emitters. There was some overlap between the groups. These two groups were
kept separate throughout the rest of the IDR analysis. (2) The total HC, CO, and NOx
emissions from all of the High emitters in the sample was calculated. (3) A total of 75 HC
/ CO cutpoint combinations were developed. These ranged from (0.5g/mi HC / 5g/mi CO)
to (5.0g/mi HC / lOOg/mi CO). For NOx, eight cutpoints were used that ranged from 1.0
g/mi to 5.0 g/mi. (4) The runningLA4 emissions identification rate (DDR) was determined
for each cutpoint combination. For example, the strict cutpoint combination of 0.5 g/mi
HC / 5.0 g/mi CO might identify 90 percent of the total emissions of the high emitters
whereas the lenient cutpoint combination of 5.0 g/mi HC / 100 g/mi CO might identify only
10 percent of the total emissions. (5) The identification rate (IDR) were calculated for 75
HC/CO cutpoint combinations, and these points were least squared regressed versus the
natural logarithms of the HC and CO cutpoint. Natural log regressions were used because
they produced better fits, and better satisfied the inherent assumptions behind least squares
linear regression. The logarithm form also makes sense physically given the skewed
distribution of emissions. For example, a change of the HC cutpoint from 1.0 to 1.5 g/mi
has a larger effect on IDR than a change from 4.0 to 4.5 g/mi. The regression coefficients
are shown in Equations 9 and 10. (6) The NOx emission identification rate (IDR) were also
calculated for eight cutpoints and fitted to a cubic equation. The cubic form was chosen
because it provides a very good fit, and does not create anomalous results such as an IDR
decrease as the cutpoint gets more stringent (See Appendix C).
In MOBILE6, the IDRs for all 1981 and later cars and light trucks are represented
by Equations 9 through 11. Where In(HCcut), In(COcut), and ln(NOcut) are the cutpoints
transformed into natural logarithm space.
HC IDR= 1.1451 - 0.1365*ln(HCcut) - 0.1069*ln(COcut) Eqn 9
COIDR= 1.1880-0.1073*ln(HCcut)-0.1298*ln(COcut) Eqn 10
M6IM001.WPD DRAFT 26 Mar 24, 1999
-------
DRAFT
The NOx IDR equation is a cubic form:
NOxIDR = 0.5453 + 0.7568*NOcut - 0.3687*NOcut2 + 0.0406*NOcut3 Eqn 11
3.10 I/MNon-Compliance Rates
One potential problem in I/M is that of non-compliant vehicles. By definition,
these are the high emitting vehicles which fail the initial test, but drop out of the I/M
process prior to receiving a passing test or a cost waiver. This type of non-compliant
vehicle is assumed to remain a high emitter at the average high emitter emission level (no
reduction is given like in the case of cost waived vehicles). In the MOBILE6 model, the
non-compliant vehicles will be represented as a fraction of the identified high emitters that
did not pass or receive a cost waiver. A default value of 15 percent will be built into the
model for the non-compliance rate. It is based on studies where large samples of I/M
vehicles were tracked as they passed through I/M programs. Optional user inputs will also
be available that permit any number from 0 percent to 100 percent to be used if supported
by data.
The other type of non-compliant vehicle is one which does not show up for its
initial test (owner ignores I/M). If these vehicles are normal emitting vehicles (passing the
I/M test) they have no effect on the result; however, if they are high emitters then they
should have the same effect as the initial failures which never pass or get waived.
Unfortunately, because they do not show up for I/M it is impossible to determine these
statistics. As an approximation, it is assumed that the 15 percent non-compliance rate
(from above) includes the effect of high emitters which did not show up for their first test.
Similarly a user defined input for non-compliance should take these vehicles into account.
3.11 Average Emissions After I/M
An important step in calculating the I/M credits is to calculate the average emissions
of the fleet after a cycle of I/M testing and repair. The average is calculated for each
vehicle type / pollutant / model year group/ technology group at many odometer points
during the life of the group. Conceptually, the average emissions of the fleet after I/M is
a weighted sum of (1) the normal emitters which were unaffected by I/M, (2) the high
emitters which were not identified by I/M and which keep the same high emissions, (3) the
high emitters which were non-compliant and which keep the same high emissions, (4) the
high emitters which were identified and cost waived, and (5) the high emitters which were
identified and successfully repaired by the I/M process. The last type drops down to the
after repair levels (FIX in Equation 12) calculated in Section 3.6 (Line D).
M6IM001.WPD DRAFT 27 Mar 24, 1999
-------
DRAFT
Equation 12 is used to calculate the average emissions of the fleet after a cycle of
I/M.
EIM = N*(1-X)+ H*X*(1-IDR)+ X*IDR*W*H*RW +
N*R*X*IDR*FIX + H*X*IDR*NC Bpl2a
Where:
N*(1-X) = Normal Emitters emission effect
H*X*(1-IDR) = High Emitters not identified emission effect
X*IDR*W*H*RW = High Emitters identified and Waived emission effect
N*R*X*IDR*FIX = High Emitters identified and Repaired emission effect
H*X*IDR*NC = High Emitters identified and Non-Compliant vehicles emission
effect
Variables:
N Emission Level of Normal Emitters (g/mi).
H Emission Level of High Emitters (g/mi).
X Fraction of High Emitters in the fleet before the cycle of I/M.
DDR Fraction of Total Fleet High Emitters Identified by an I/M test.
W Fraction of Identified High Emitters which get a repair cost waiver.
NC Fraction of Identified High Emitters which are in non-compliance of the I/M
program.
FIX Fraction of Identified High Emitters which get repaired to pass the test.
R Fraction of the normal emitter level that high emitters are repaired after I/M
(value is > 1.0).
RW Fraction of the high emitter level that waived high emitters are repaired
after I/M.
The fractions W, NC, and FIX are all applied to the DDR fraction. An identified
vehicle is either waived, in non-compliance, or is properly repaired. Thus,
W + NC + FIX = 1.0 BpEb
M6IM001.WPD DRAFT 28 Mar 24, 1999
-------
DRAFT
4.0 Periodic I/M Credit Algorithm (Sawtooth Methodology)
This section describes the methodology for the periodic I/M credit algorithm over
time or the "sawtooth" methodology. The first few sections describe the algorithm and
equations for the "sawtooth" when neither a remote sensing program that identifies high
emitters, nor a change of ownership testing requirement is present. The remaining sections
describe the "sawtooth" algorithm for a combined program of periodic I/M (I/M), change
of ownership I/M (COIM), and remote sensing. Both algorithms are essentially the same,
except the I/M + RSD + COIM algorithm works on a bi-annual (every six months) basis,
and the I/M only algorithm works on an annual or biennial basis. MOBILE6 will also
compute benefits for other variations of I/M, RSD, and COIM. These include: I/M+RSD,
I/M+COEVI, and COIM only. The algorithm for these variations is essentially the same as
for the base I/M+RSD+COEVI case.
4.1 Discrete Points
The MOBILE6 program will not use "continuous" regression lines of emissions
versus mileage to represent the before and after I/M emission rates, but instead will use
discrete points on these lines. Each point on the line will represent a particular vehicle age
that ranges from 1 to 26 years. Table 3 shows the correspondence of age and cumulative
mileage for cars. Each particular age and mileage corresponds to a January 1st reference.
The six month mileage points needed for RSD and COIM will be generated from the
mileages on this table by averaging the two surrounding points. For example, age =1.5
years (18 months) is obtained by averaging the points at age = 1 and age = 2.
The text describing the I/M credits with NO RSD or COIM will use the index
variable 'ii' to represent yearly intervals. The text describing the I/M credits with RSD
and/or COIM will use the index variable 'i' to represent six month intervals.
M6IM001.WPD DRAFT 29 Mar 24, 1999
-------
DRAFT
Table 3
January 1st Mileage and Age Correspondence for Cars in MOBILE6
Age
1
2
3
4
5
6
7
8
9
10
11
12
13
Cumulative Mileage
(in 1000's)
2.142
12.823
29.335
45.050
60.006
74.239
87.786
100.678
112.948
124.625
135.738
146.315
156.380
Age
14
15
16
17
18
19
20
21
22
23
24
25
26
Cumulative Mileage
(in 1000's)
165.960
175.077
183.753
192.010
199.869
207.349
214.466
221.241
227.688
233.823
239.663
245.220
250.509
4.2 Effect of Exemptions on I/M Credits
I/M exemptions are a provision granted to some vehicles which would ordinarily
be subject to an I/M inspection that excuses them from all of the testing and repair
requirements of I/M. In practice, this means that the motorist does not have to bring the
vehicle in for an I/M test; however, it may require the motorist to have received a roadside
remote sensing device (RSD) "clean screening" test(s), or to have paid a fee in-lieu of the
test.
MOBILE6 will be able to account for consecutive age / model year exemptions
starting with age = 1. For example, most programs exempt the youngest fleet age or the
newest model year in the fleet. This means that vehicles which are one year old are not
tested. Because it is so common, a one year exemption is the default pattern shown for the
M6IM001.WPD DRAFT
30
Mar 24, 1999
-------
DRAFT
annual and 1-3-5 biennial I/M algorithms. A two year exemption is the default pattern for
the 2-4-6 biennial algorithm. MOBILE6 will also have the ability to exempt the first of any
consecutive new model years (i.e., exempt vehicles 2, 3, 4, 5, ... n years old). The effect
of this is to shift the 'sawtooth' curve to the right. Exempting older model year / age
vehicles (i.e., all of the vehicle 15 years or older) will also be available in MOBILE6.
Mechanically, this will be done by computing the credits using the standard algorithm, and
then setting them to zero for the exempted model years.
4.3 Annual I/M Credits with NO Remote Sensing
The MOBILE6 model will generate separate I/M credits for each combination of
vehicle type / pollutant / model year group / technology class. In concept, these credits
could be generated by comparing the basic emission rate line - (see Section 3.3) with the
average emissions after I/M line - (see Section 3.9). However, because of a number of
complications these lines cannot be used directly. Instead EPA developed the 'sawtooth'
algorithm shown conceptually in Figure 2a.
One of the problems with a linear approach is the distribution of ages within a
model year group. For purposes of modeling, all vehicles are assumed to be inspected on
the first anniversary of their purchase and periodically thereafter, always on that same date.
It is also assumed that sales occur exactly in the 12 month period from October of the
calendar year previous to the model year through September of the next calendar year. For
example, in January, 1999, the age distribution of the 1997 model year vehicles will range
from 2.25 years to 1.25 years. With an annual inspection program, vehicles between one-
and-two-years-old have only been inspected once. Any vehicles two years and older should
have received their second inspection. In this example, 25 percent of the emissions on the
evaluation date come from vehicles recently completing their second inspection and 75
percent of the emissions come from vehicles which have been inspected only once.
Another factor which must be taken into account is the deterioration of the vehicles
in between their yearly inspections and repairs. Existing evidence suggests that the type
of problems which cause I/M failures can re-occur as often in the repaired vehicles as they
do in the unrepaired fleet. Thus, it is assumed that the fleet, after repair, will have the same
emission deterioration as before repairs.
Figure 2a graphically shows the I/M credit methodology. The top set of 26 (ages)
points (only 6 are shown) is the basic emission rate for a given group (vehicle type /
pollutant / model year group / technology). For instance, Points B, C, and D show the non
I/M line for vehicle ages of 1, 2, and 3 years. The lower 'sawtooth' figure is the I/M line.
The 'sawtooth' illustrates the effect of I/M inspection and repair and the subsequent
M6IMOO1.WPD DRAFT 31 Mar 24, 1999
-------
DRAFT
FIGURE 2A
ANNUAL I/M CREDITS SAWTOOTH
LA4
EMISSIONS
B
MILEAGE
deterioration of the fleet. All deterioration slopes are parallel (i.e., segment B-C is parallel
to segment E-F). The repair effect is represented by the sudden drop in emission level at
each inspection interval (i.e, from Point F to Point G). The heavy shaded portions of the
lines illustrate how an I/M credit for the given group at age X is produced. MOBILE6
always chooses January 1 st as the evaluation date. The vehicles sold from October through
December are represented by the short line segment to the right of the two year anniversary
point. These are vehicles which are older than X years. The longer line segment to the left
of the anniversary point represents the vehicles sold from January through September,
which are still less than X years old at the January evaluation date. The weighted average
of each segment is calculated and the percent difference between the two weighted averages
is computed. This percent difference is the I/M credit for a given age.
Mathematically, this process is shown for the Non I/M (top line) and the I/M
(sawtooth line) as:
M6IM001.WPD DRAFT
32
Mar 24, 1999
-------
DRAFT
Nonl/M
NOIM(ii) = 0.75 * (E(ii) - 0.375*(E(ii) - E(ii-l)) ) +
0.25 * ( E(ii) + 0.125*(E(ii) - E(ii+l)) ) Egn 14
Where NOIM is the average Non I/M value at age = ii in the equations above. It is
the average of the two segments. Note "ii" means the "ii th" point. It should not be
confused with Point I on the Figure 2B.
E(ii) is the basic emission rate at point ii. E(0) is the value at Point A.
The values of E(ii), E(ii-l), and E(ii+l) take into account the slope of the line between age
= ii and age = ii + 1 and age = ii -1. Figure 2a is an idealized drawing using a straight line.
In the MOBILE6 model, the lines have some slight curvature due to the high emitter
correction factor; thus, the slope is generally not the same between all segments.
Also, the weighting factor values of 0.375 and 0.125 shown throughout these equations
represent the average of the heavy shaded segments in Figure 2A. For example, the 0.375
represents the average weighting of the highlighted segment between points B and C, and
the 0.125 represents the average weighting of the highlighted segment between points C
andD.
I/M Special Case age = ii = 1
IM(1) = 0.75 * ( E(ii) - 0.375*(E(ii) - E(ii-l) ) +
0.25 * ( EIM(ii) + 0.125*(E(ii + 1) - E(ii)) ) Egn 15
General Case age > 1
IM(ii) = 0.75 * [ EIM(ii) - 0.375*(E(ii) - E(ii-l) ] +
0.25 * [ EIM(ii) + 0.125* (E(ii + 1) - E(ii)) ] Egn 15
Where EVI(ii) is the I/M line. Where EIM(i) is the average I/M emission line after
repair, waiver, and non-compliance factors from Section 3.9.
The I/M credits are computed by dividing the difference between the NOIM
emission value and the EVI emission value by the NOIM emission value. An I/M credit is
obtained for the ages 1 through 25.
IMCRED(ii) = [NOIM(ii) - IM(ii)] / NOIM(ii) Egn 15
M6IMOO1.WPD DRAFT 3 3 Mar 24, 1999
-------
DRAFT
4.4 Biennial I/M Credits with NO Remote Sensing
The values of E(ii) and EIM(ii) used in Equations 13 through 16 are also used to
compute biennial I/M credits using a 'sawtooth' algorithm. The only difference between
the annual and the biennial I/M credits is that the biennial values are applied every other
year and that there is consequently a longer period of deterioration between I/M inspections
and repairs. Figures 2b and 2c are analogous to Figure 2a. Figure 2b is an example of a
1-3-5 biennial program in which a vehicle is first inspected when it is one year old and then
every two years thereafter. Figure 2c illustrates a 2-4-6 biennial program which first
inspects a vehicle when it is two years old and then does an inspection every other year.
The differences are small for a fleet that has a full complement of vehicle ages. The
"Mixed" Biennial credits (Mix Bien EVICRED) are an average of these two program types.
"Mixed Biennial" was the default for MOBILES. This is an average of the 1-3-5 and 2-4-6
plans, or any mixed biennial program in which half of each model year or half of the fleet
is inspected during each calendar year.
M6IM001.WPD DRAFT 34 Mar 24, 1999
-------
FIGURE 2B
1-3-5 BIENNIAL I/M CREDITS SAWTOOTH
LA4
EMISSIONS
MILEAGE
FIGURE 2C
2-4-6 BIENNIAL I/M CREDITS SAWTOOTH
LA4
EMISSIONS
MILEAGE
M6IM001.WPD DRAFT
35
Mar 24, 1999
-------
DRAFT
Years of exemption = X. The Biennial 1-3-5 pattern and the Biennial 2-4-6 pattern
have the same general equations except i starts at i=l (X=l) for the first and i=2 for the
second (X=2).
Nonl/M
NOIM(ii) = 0.75 * [E(ii) - 0.375*(E(ii) - E(ii-l)) ] +
0.25 * [E(ii) + 0.125*(E(ii) - E(ii+l)) ] Eqn 14
I/M Special Case ii = 1 for 1-3-5 Case or ii = 2 for 2-4-6 Case
IM(1) = 0.75 * [ E(ii) - 0.375*(E(ii) - E(ii-l) ] +
0.25 * [ EIM(ii) + 0.125*(E(ii + 1) - E(ii)) ] Eqn 15
General Case ii > 2*X
IM(ii) = 0.75*[EIM(ii-2)+E(ii-l)-E(ii-2) + 0.625*(E(ii)-E(ii-1))]
0.25* [EIM(ii) + 0.125*(E(ii + 1)-E(ii)) ] Eqn20
IMCRED(ii) = [NOIM(ii) - IM(ii)] / NOIM(ii) Eqn21
MixBien IMCRED = [(1-3-5 Bien Credit+ 2-4-6 Bien Credit] / 2 Eqn22
4.5 I/M Credits with Remote Sensing and Change of Ownership
The MOBILE6 program will be able to calculate I/M credits for programs which
conduct periodic I/M tests, RSD testing to identify high emitters, and change of ownership
(COIM) I/M credits. The methodology in Sections 4.3 and 4.4 was for periodic I/M only
type programs. Since RSD and COIM are non-periodic in nature, it is assumed that they
identify and repair vehicles in a continuous distribution throughout a calendar year (i.e., all
of the RSD testing or change of ownership testing is not conducted in June). Based on this
assumption, this testing pattern is equivalent to a periodic test every six months. The RSD
factor used in Equations in Sections 4.5.1 through 4.5.3 is computed as a product of the
M6IMOO1.WPD DRAFT 3 6 Mar 24, 1999
-------
DRAFT
RSD effectiveness parameter (Effectiveness), and the RSD Fleet coverage (Coverage).
These two basic parameters account for the RSD's ability to individually identify high
emitters, and its ability to test the entire fleet.
Also, the RSD and COEVI credits are computed at each six month interval between
the regular periodic inspections. Thus, there is one RSD / COEVI interval (sawtooth
pattern) between each periodic inspection in an annual program, and three RSD / COEVI
intervals between each biennial inspection. This assumes that on average the entire I/M
fleet will not be inspected by RSD more than once during a year.
4.5.1 RSD Effectiveness Values
Table 4 shows the RSD effectiveness for HC, CO and NOx pollutants versus RSD
percent CO readings. The individual values in the table represent the emissions identified
by RSD at particular cutpoints ranging from 0.5% CO to 7.5% CO. The effectiveness
values are based on studies done by EPA in Arizona, and by CARB in Sacramento. These
values will be used as the RSD Effectiveness parameters in the MOBILE6 model. For
more details on how these values were derived see "EPA420-R-96-004 - "User Guide and
Description for Interim Remote Sensing Program Credit Utility".
M6IMOO1.WPD DRAFT 3 7 Mar 24, 1999
-------
DRAFT
Table 4
Remote Sensing Effectiveness Versus CO Cutpoint for 1981-93 Model Year Vehicles
RSD CO Cutpoint
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
4.5%
5.0%
5.5%
6.0%
6.5%
7.0%
7.5%
HC Effectiveness
0.570
0.433
0.387
0.348
0.319
0.262
0.217
0.182
0.150
0.109
0.071
0.060
0.046
0.039
0.028
CO Effectiveness
0.596
0.499
0.442
0.396
0.352
0.278
0.213
0.178
0.133
0.107
0.072
0.053
0.044
0.034
0.017
NOx Effectiveness
0.283
0.178
0.122
0.091
0.059
0.054
0.042
0.018
0.015
0.009
0.006
0.003
0.003
0.003
0.003
4.5.2 RSD Coverage Options
Three basic RSD program options will be available to the MOBILE6 user. Option
1 is the "Level of Effort Commitment", Option 2 is the "Specific Level of Fleet Coverage
Commitment", and Option 3 is the "Number of Failures Commitment". All three of these
options utilize the coverage and effectiveness parameters. However, different methods are
used to calculate the two parameters, and different user inputs are required.
(See RSD references for more details).
M6IM001.WPD DRAFT
38
Mar 24, 1999
-------
DRAFT
Option 1 - Level of Effort Commitment
In this option, the user enters the number of vehicle RSD tests which are to be done
on an annual basis, and the total size of the fleet (in number of vehicles) which is subject
to inspection. A modified Poisson algorithm is used to estimate the number of vehicles
seen by remote sensing in order to calculate the fraction of the fleet tested by RSD
(Coverage). This is necessary, since the fraction of all vehicles in the fleet which are
measured by remote sensing is a function of the total number of RSD readings since some
vehicles are seen multiple times by RSD. A Poisson algorithm is a standard method to
model such a situation. The coverage fraction is also a function of the annual average
VMT of a vehicle model year at a given age compared to its VMT when new. The
equations which are used are:
P(X)
Lambda* *X * exp(-Lambda) / X!
Where,
X
P(X)
Lambda
Lambda =
Is an integer number starting at zero that represents the number of
RSD tests which a vehicle receives before it is called in for a
confirmatory test. Most programs will require at least two failing
tests (X=2) prior to the confirmatory test.
The vehicle age under evaluation.
The Poisson distribution function
Is the mean number of RSD tests during a given year. For example,
if half of the fleet is inspected on average then Lambda would be
0.50. Mathematically, it is represented as:
# RSD tests/yr * VMT(i) / ((# Veh in Fleet)*VMT(l))
Coverage(X) = 1.0 - SUM(P(X-1))
"No RSD" is the default option, However, if RSD is to modeled then Option 1 is
recommended.
M6IM001.WPD DRAFT
39
Mar 24, 1999
-------
DRAFT
Option 2 - Specific Level of Fleet Coverage Commitment
In this option, the user specifies the fraction of the fleet in each model year that is
seen using remote sensing. This fraction should only be the fraction of the fleet which has
had sufficient valid remote sensing measurements to be identified as remote sensing
failures for purposes of further I/M inspection. For example, if the RSD program is
designed so that three RSD failures are needed before a vehicle is sent for off-cycle I/M
testing (confirmatory testing), the fraction of the fleet used as the coverage commitment
should represent the portion of the fleet that has received three RSD readings.
Thus,
Coverage = User Input
Option 3 - Number of Failures Commitment
In this option, the user specifies the fleet size by model year (in units of number of
vehicles), and number of confirmed I/M failures by model year which the RSD will
identify. Only vehicles identified for inspection by remote sensing and which fail the I/M
inspection are to be included. In the MOBILE6 model, this option effectively combines the
Effectiveness and the Coverage into one parameter - the fraction of high emitters identified
by RSD. This value is then normalized to the entire fleet fraction of high emitters. The
variable "Highs" from Equation 7 in Section 3.8 is used to normalize the RSD failure rate.
The normalized value becomes the RSD parameter used in the equations in Sections 4.5.2
and 4.5.3.
RSD Fail Rate = # RSD Failures / Fleet Size
RSD = RSD Fail Rate / Highs
4.5.3 RSD Effective and Coverage Together
The final RSD value used in the I/M + RSD credit equations in Equations 4.5.2 and
4.5.3 is the product of the RSD Effectiveness and the RSD Coverage. For Coverage
Options 1 and 2, the values are calculated separately and multiplied together. For Coverage
Option 3, the RSD effectiveness and coverage are implied in the RSD Fail Rate. To assure
consistency between all three coverage methods, MOBILE6 will prompt the user if
unreasonable values for "RSD" or RSD Coverage parameters are used.
RSD = RSD Coverage * RSD Effectiveness
M6IM001.WPD DRAFT 40 Mar 24, 1999
-------
DRAFT
4.5.4 Change of Ownership Parameters
The change of ownership testing frequency is assumed to be quite random, and will
be modeled for simplicity by the same six month interval as RSD. This effectively assumes
that the change of ownership vehicles are representative of the overall fleet, and that this
fraction of the fleet receives an additional I/M test every six months. Based on IM240 data
from Wisconsin, this fraction was determined to be 16 percent annually. A default value
of 8 percent will be applied for each six month interval. The 16 percent value was
estimated from the Wisconsin change of ownership versus periodic test volume data shown
in Table 5. The numbers are test vehicle counts, and are based on a quasi-random sample
of full IM240 initial tests conducted either at Station #12 in Wisconsin or at other
Wisconsin test stations on Saturdays. The Wisconsin I/M program is biennial and the
periodic tests for the 1996 calendar year are on the odd model years only whereas the
change of ownership testing is on all model years.
In addition, anecdotal evidence from Wisconsin suggests that this model may be a
simplistic, and under predict the benefits of change of ownership. For example, analysis
of change of ownership vehicles suggests that they (1) often contain a higher percentage of
high emitters than the overall fleet, (2) that the high emitters change ownership more
frequently than the more normal emitters, and (3) that a percentage of the change of
ownership vehicles change owners more than once during a year. Therefore, to help
balance these possible effects, the effect of waivers and non-compliance will not be
assessed on change of ownership I/M testing.
Because of the uncertainty in estimating change of ownership parameters, the user
will be allowed to input their own value into the MOBILE6 model. The COIM factor used
in Equations in Sections 4.5.1 through 4.5.3 is computed as a product of the change of
ownership fraction and the high emitter identification rate (IDR). These two factors
account for how many change of ownership tests are done, and the effectiveness of each
test.
M6IMOO1.WPD DRAFT 41 Mar 24, 1999
-------
DRAFT
Table 5
Wisconsin Change-of-Owner Test Volumes
in a 1996 Calendar Year Data Sample
Model Year
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
ALL
Periodic Testing
100
275
679
943
1231
1238
4,466
Change Owner
Testing
31
48
62
201
147
250
212
280
209
250
181
842
% Change of
Owner
24.0%
18.4%
17.8%
18.4%
14.5%
12.8%
15.9%
The RSD / COIM credits are computed as a fraction of the maximum possible
periodic I/M credit at that given age. This has the effect of producing smaller "sawteeth"
for six month intervals which do not coincide with a periodic inspection (smaller means
that the bottom of the "sawtooth" is higher than the bottom of the periodic inspection
"sawtooth"). There is no additional credit given, when the RSD / COIM intervals coincide
with the periodic inspection interval. See Figures 3a and 3b.
4.5.1 Special Case for Year 1 of Program
Exemption = 1 Year (Annual I/M or 1-3-5 Biennial) then i = 2
Exemption = 2 Years (2-4-6 Biennial) then i = 4
M6IM001.WPD DRAFT
42
Mar 24, 1999
-------
DRAFT
Exemption = X Years then i = 2*X
Index variable 'i' represents six month intervals; thus, 2*i = ii
Nonl/M
NOIM(i) = 0.75 * ( E(i) - 0.375*(E(i) - E(i-l)) ) +
0.25 * ( E(i) + 0.125*(E(i) - E(i+l)) ) Eqn 14
I/M
(a) Points E(i) and EIM(i) are all known.
(b) I/M and non-I/M lines are parallel
EO/2) = (E(i-2HE(i-lV)
2
E(i-2), E(i-l), E(i), and E(i+l) are Known Values
EEVI(i), EEVI(i+1) are Known Values
TOP(i+l) = EIM(i) + (E(i+l) - E(i) )
SEGMENT 1A = (Ef1/^ + Ei
SEGMENT2A= (EIM(i) + [EIM(i) + (E(i+1) - E(i))/2])
2
IM(i) = 0.75 * SEGMENT1A + 0.25 * SEGMENT 2A
IMCRED(i) = fNOIM(i) - I
JNUlM(l)
4.5.2 Annual I/M Credits
M6IM001.WPD DRAFT 43 Mar 24, 1999
-------
DRAFT
General Case#l: Annual I/M for Year 2 through 25;
General algorithm is i + n
RSD = RSD Coverage * RSD Effectiveness
COEVI = Change of Ownership Fraction * IDR
RSD + COIM<=1.0
E (i - 2)....E (i + 1) are derived from the basic emission factors
EEVI (i - 2)....EEVI(i + 1) are calculated in Section 3.11.
TOP (i - 1) = EIM(i - 2) + (E (i-1) - E(i - 2))
MID (i - 1) = TOP(i-l) - [TOP(i - 1) - EIM(i - 1)] * RSD
TOP(i) =MID(i- 1) + (E(i) - E (i - 1) )
MID(i) = EIM(i)
TOP(i + 1) = MID(i) + (E (i + 1) - E(i))
SEGMENT IB =
I \EIM (i - 2) + (E(i -1) - E(i - 2)) + TOP (i -1) I/ 2
= (jVIID (i -1} + TOP (i))
SEGMENT 2B = \ MID (i - 1) + TOP (i)
2
SEGMENT 3B = I MID(i) + [ MID(i) + ECi +2D -E(D J I / 2
IM(i) = 0.25 * SEGMENT IB + 0.50 * SEGMENT 2B + 0.25 * SEGMENT 3B
IMCRED(i) = (NOIM(n - IM(D
NOIM( i)
M6IM001.WPD DRAFT 44 Mar 24, 1999
-------
DRAFT
4.5.3 Biennial I/M Credits
General Case #1: Biennial I/M for Year i through 25;
General algorithm is i + n
i = 2 for the 1-3-5 Biennial
i = 4 for the 2-4-6 Biennial
CHART
General
i-4
i-3
i-2
i- 1
i
i+ 1
Example
2
3
4
5
6
7
RSD = RSD Coverage * RSD Effectiveness
COEVI = Change of Ownership Fraction * DDR
RSD + COIM<=1.0
E (i - 2)....E (i + 1) are derived from the basic emission factors
EEVI (i - 2)....EEVI(i + 1) are calculated in Section 3.11.
EIM(i-4) = MID(i-4)
TOP(i-3) = MID(i-4) + (E(i-3) - E(i-4))
MID(i-3) = TOP(i-3) - [TOP(i-3) - EIM(i-3)] * RSD
TOP(i-2) = MID(i-3) + (E(i-2) - E (i-3))
MID(i-2) = TOP(i-2) - [TOP(i-2) - EIM(i-2)] * RSD
TOP(i-l) = MID(i-2) + (E(i-l) - E(i-2))
MID(i-l) = TOP(i-l) - [TOP(i-l) - EIM(i-l)] * RSD
TOP(i) =
EIM(i) = MID(i)
TOP(i+l) =
M6IM001.WPD DRAFT 45 Mar 24, 1999
-------
DRAFT
SEGMENT IB =
I (EIM (i - 4) + ( E(i - 3) - E(i - 4))/2 + TOP (i - 3) I/ 2
= (j
SEGMENT 2B = \ MID (i - 3) + TOP (1-2)
2
SEGMENT 3B =
I MID(i-2) + [ MID(i-2) + (E(i-l) - E(i-2))/2 J I / 2
IM(i) = 0.25 * SEGMENT IB + 0.50 * SEGMENT 2B + 0.25 * SEGMENT 3B
IMCRED(i) = (NOIM (i 1 - IMCD
NOIM( i)
I (jVHD(i-2) + ( E(i-l) - E(i-2))/2 + TOP (i - 1) I/ 2
SEGMENT 1C =
SEGMENT 2C = \MSD (i -1) + TOP (i)J
2
I MID(i) + |_ MID(i) + (E(i+l) - E(i))/2J I / 2
SEGMENT 3C = I MID(i) + MID(i) + (E(i+l) - E(i
IM(i+l) = 0.25 * SEGMENT 1C + 0.50 * SEGMENT 2C + 0.25 * SEGMENT 3C
IMCRED(i+l) = (NOIM (i 1 - IM(i1
NOIM( i)
M6IM001.WPD DRAFT 46 Mar 24, 1999
-------
DRAFT
4.6 RSD / Vehicle Profiling Exemptions
4.6.1 RSD Exemptions
An important use of the RSD test in the context of I/M may be to use it as a method
of screening out 'normal' emitting vehicles, and exempting them from regular I/M. The
motivation for such a program might be to reduce the inspection cost by exempting a
fraction of the fleet which is very likely to pass anyway. Also, since the test is largely
unknown to the vehicle owner, and rather automatic, it might help build public support for
an I/M program by inconveniencing fewer motorists.
The RSD clean screening logic is similar to that used in the high emitter
identification algorithm. Both involve the terms RSD fleet coverage and RSD
effectiveness. However, clean screening is attempting to properly identify low emitting
vehicles for exemption from further program requirements while RSD high emitter
identification is concerned with identifying high emitters for further testing and repair.
Clean Screening Coverage Options
Options 1 and 2 presented in Section 4.5.2 will be used for the clean screening
coverage. These two options are the "Level of Effort Commitment", and the "Specific
Level of Fleet Coverage Commitment". The same equations and algorithms will be used
to model clean screening coverage as were used to model high emitter identification
coverage (i.e., Poisson distribution equations). Option 3 will not be used because it only
applies to high emitters.
Clean Screening Effectiveness Values
The RSD clean screening effectiveness values are shown in Tables 6a and 6b. They
are shown as a percentage of the I/M credit which is lost through clean screening. They are
a function of the RSD cutpoint combination, and the stringency of the underlying I/M test
from which a clean screened vehicle is exempted. Also shown in the table is the percentage
of the fleet which is exempted (clean screened). For example, if the RSD cutpoints of 200
ppm HC and 0.5% CO are used for clean screening, and the less stringent (phase-in) EVI240
cutpoints are used, then 51 percent of the RSD tested fleet is exempted, 2 percent of the HC
IDR is lost, 7 percent of the CO IDR is lost and 23 percent of the NOx IDR is lost.
M6IM001.WPD DRAFT 47 Mar 24, 1999
-------
DRAFT
RSD Clean Screening Effectiveness and Coverage Together
The RSD clean screening effectiveness and RSD coverage are multiplied together
to produce an overall RSD clean screening effect. Mathematically, this is:
RSD Loss
RSD Coverage * RSD Clean Screening Effectiveness
The resulting value for the RSD_Loss is applied by subtracting it from the I/M DDR
(IDR) obtained in the previous I/M credit equations in Section 3.8. This produces the final
IDR_RSD. For example, if the original I/M IDR (IDR) is 80 percent, the RSD coverage
is 50 percent, and the losses from falsely exempting high emitters using RSD is 2 percent,
then the I/M credit with RSD (IDR_RSD) is 79 percent.
A simplified mathematical equation is:
IDR RSD
IDR - RSD Loss
Table 6a
Remote Sensing Clean Screening Effectiveness
Interim (Less Stringent) I/M Standards
Clean
Screening
Outpoints
HC 200 ppm
CO 0.5%
NOx - None
HC 200 ppm
CO 0.5%
NOx - 2000 ppm
HC 200 ppm
CO 0.5%
NOx- 1500 ppm
HC 200 ppm
CO 0.5%
NOx- 1000 ppm
Vehicles
Tested
594
594
594
594
% Vehicles
Passing Clean
Screening
51%
40%
37%
29%
% HC IDR
Credit LOST
2%
2%
1%
1%
% CO IDR
Credit LOST
7%
7%
0%
0%
% NOx IDR
Credit LOST
23%
12%
11%
7%
The default RSD exemptions used in the MOBILE6 model are based on an
extensive study of RSD data and I/M data collected in various cities. The full methodology
M6IM001.WPD DRAFT
48
Mar 24, 1999
-------
DRAFT
and numbers used in the MOBILE6 model are fully documented in the EPA document
"Draft Description and Documentation for Interim Vehicle Clean Screening Credit Utility -
EPA420-P-98-008".
Table 6b
Remote Sensing Clean Screening Effectiveness
Final (More Stringent) I/M Standards
Clean
Screening
Outpoints
HC 200 ppm
CO 0.5%
NOx - None
HC 200 ppm
CO 0.5%
NOx - 2000 ppm
HC 200 ppm
CO 0.5%
NOx- 1500 ppm
HC 200 ppm
CO 0.5%
NOx- 1000 ppm
Vehicles
Tested
594
594
594
594
% Vehicles
Passing Clean
Screening
51%
40%
37%
29%
% HC IDR
Credit LOST
9%
6%
5%
4%
% CO IDR
Credit LOST
7%
5%
1%
1%
% NOx IDR
Credit LOST
28%
15%
12%
7%
4.6.2 High Emitter Profiling
High emitter profiling is similar to RSD in that it seeks to screen out low emitting
vehicles and exempt them from the regular I/M inspection. The benefit is a saving of
testing resources, and less inconveniencing of motorists. Like RSD the drawback is the
loss of I/M benefits from exempting high emitters which should not be exempted. The
equations which are used (shown below) are completely analogous to the RSD equations
in terms of form and use.
Profile_High = Function [ %fleet Profiled, Error Rate of Profile]
IDR_Prof = IDR - Profile_High
A user of MOBILE6 may want to model an I/M program which does both RSD and
high emitter profile exemptions. In that case, both the RSD_High and the Profile_High
losses are subtracted from the based IDR to produce a new IDR.
IDR RSD Prof
IDR - RSD_High - Profile_High
M6IM001.WPD DRAFT
49
Mar 24, 1999
-------
DRAFT
5.0 I/M ALGORITHM FOR START EMISSIONS
5.1 General I/M Algorithm
The MOBILE6 model will also compute I/M credit reductions for start emissions
in addition to the running LA4 emissions. The start I/M credits will be small in magnitude
since the typical I/M test (i.e., EVI240, idle, etc) does not intentionally involve testing a
vehicle during start or warm-up. The I/M credits for start emissions will reflect this fact
by assuming that vehicles with high start emissions are identified in conjunction with a
running emission failure.
The generalized structure of the start I/M credit algorithm is the same structure as
used for the running LA4 emission credits (See Figure 1). However, the Y-axis represents
start emissions in grams and the X-axis represents mileage. Line A shows the basic start
emission factor line before an I/M reduction. Line B shows the average start emissions of
the normal emitting vehicles. Line C shows the average start emissions of the high emitting
vehicles.
5.2 I/M Start Emission Rates
The basic emission rates for start emissions (Line A of Figure 1) and the
methodology used to develop them can be found in the EPA document "Determination of
Start Emissions as a Function of Mileage and Soak Time for 1981-1993 Model Year Light-
Duty Vehicles" - Report Number M6.STE.003.
Table 4 contains the start emission regression coefficients for the normal emitting
vehicles for all eight technology and model year groups. Table 5 contains the average start
emissions from the high emitting vehicles (high emitters are defined based on twice or
thrice FTP standards - see Section 3.2). Table 6 shows the average after repair level of the
high emitting vehicles. The values shown in Table 6 are based on after repair emission
testing. In these cases high emitting vehicles (high FTP emissions or EVI240 failures) were
tested, repaired and retested. The analysis of the start emissions before and after repair is
discussed in detail in EPA document M6.IM.002 "Determining Repair Effects of EVI240
Cold Start Emissions for 1981 and Later Light-duty Vehicles".
M6IM001.WPD DRAFT 50 Mar 24, 1999
-------
DRAFT
Table 4a
Regression Coefficients for START Emissions from Normal Emitter CARS
MY
Group
1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group
PFI
TBI
FI
Carb
Carb
FI
Carb
HC Coefficients
ZML
1.9987
1.9019
2.3589
1.4934
1.5892
2.3543
2.1213
DET
0.006830
0.002679
0.001388
0.018238
0.009408
0.008533
0.013610
CO Coefficients
ZML
18.972
19.233
19.949
24.698
24.442
20.038
28.637
DET
0.00703
0.00000
0.00000
0.10947
0.10577
0.22673
0.22673
NOx Coefficients
ZML
1.444
2.300
1.461
1.405
0.748
1.530
1.601
DET
0.00220
0.00000
0.00141
0.00000
0.00524
0.00059
0.00000
Table 4b
Mean START Emissions of Hish Emitter CARS
MY Group
1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group
PFI
TBI
FI
Carb
Carb
FI
Carb
HC Mean
4.829
3.293
5.313
10.520
10.520
5.313
10.520
CO Mean
38.06
27.16
65.31
92.82
92.82
92.82
92.82
NOx Mean
Same as Normals
Same as Normals
Same as Normals
Same as Normals
Same as Normals
Same as Normals
Same as Normals
M6IM001.WPD DRAFT
51
Mar 24, 1999
-------
DRAFT
MY
Group
1988-93
1988-93
1981-87
1984-93
1981-83
Table 5a
Regression Coefficients for START Emissions from
Normal Emitter Light Trucks
Tech
Group
PFI
TBI
FI
Carb
Carb
HC Coefficients
ZML
2.873
4.073
2.599
3.916
6.817
DET
0.00000
0.01309
0.00964
0.00854
0.00154
CO Coefficients
ZML
32.178
42.456
23.497
78.286
98.432
DET
0.0168
0.1411
0.0613
0.2564
0.3240
NOx Coefficients
ZML
1.597
4.294
1.384
0.143
1.082
DET
0.00000
0.00324
0.00000
0.00436
0.00000
Table 5b
Mean START Emissions of Hish Emitter Trucks
MY Group
1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group
PFI
TBI
FI
Carb
Carb
HC Mean
5.212
5.212
5.826
9.406
17.865
CO Mean
83.862
83.862
60.319
162.115
179.549
NOx Mean
Same as Normals
Same as Normals
Same as Normals
Same as Normals
Same as Normals
M6IM001.WPD DRAFT
52
Mar 24, 1999
-------
DRAFT
Table 6
START Emission Regression Coefficients for High Emitters After Repair
Cars and Trucks
MY
Group
1990-93
1990-93
1986-89
1986-89
1983-85
1983-85
1981-82
1981-82
Tech
Group
PFI
TBI
FI
Carb
FI
Carb
FI
Carb
HC Coefficients
ZML
2.60
2.60
3.11
3.11
2.70
2.70
2.70
2.70
DET
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
CO Coefficients
ZML
18.90
18.90
30.05
30.05
28.33
28.33
28.33
28.33
DET
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.0000
NOx Coefficients
ZML
1.48
1.48
1.49
1.49
1.84
1.84
1.84
1.84
DET
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
5.3 Fraction of High and Normal Emitters in the Fleet
The basic start emission factor is computed from a weighted average of the highs
and normals. The fraction of high emitters (fraction of normal emitters = 1 - fraction of high
emitters) in the fleet is the weighting factor. The fraction of high start emitters is the same
fraction as the one used for the running emissions calculations. Tables 3a and 3b and
Appendix A in EPA document M6.STE.003 "Determination of Start Emissions as a
Function of Mileage and Soak Time for 1981-1993 Model Year Light-duty Vehicles" show
and explain the fraction of HC and CO high emitters in the fleet at selected mileages / ages
for each pollutant. The fraction of NOx high emitters is not shown because for NOx the
Normals and Highs are assumed to have the same emission rate (no start NOx highs are
assumed to exist).
M6IM001.WPD DRAFT
53
Mar 24, 1999
-------
DRAFT
5.4 I/M Start Identification Rates
The algorithm for start emissions is based on test data that indicates that a portion
of the vehicles with high running emissions that are identified by the I/M process will also
have high start emissions, and that these will be identified and corrected in conjunction
with the repairs to pass the I/M test. Also, because significant NOx emissions usually form
only after the vehicle is warm, it was assumed that an I/M program could only reduce HC
and CO start emissions.
A mathematical function that relates HC / CO cutpoint with the start emissions
identification rate (IDR) was developed from the 910 vehicle sample used to develop the
running emissions DDR. The same methodology was used to develop the Start emission
IDR as was used to develop the running emission IDR (See Section 3.9 for a more detailed
explanation). This function also has the same range of HC and CO cutpoints (HC ranges
from 0.50 g/mi to 5.0 g/mi and CO ranges from 5.0 g/mi to 100 g/mi) used in the running
emission analysis. It predicts the percentage of start emissions from high emitters which
are identified at a specific HC/CO cutpoint level. This is the percentage of the emissions
from high emitters at Line C in Figure 1 that are reduced down to average fleet emission
levels (Line A in Figure 1). The statistical results are shown in Appendix D. The functions
are:
StartHC IDR = 0.9814 - 0.1590*ln(HCCUT) - 0.1409*ln(COCUT) Eqn32
StartCO IDR= 1.1460 - 0.1593*ln(HCCUT) - 0.1707*ln(COCUT) Eqn33
5.5 Average Start Emissions After I/M
The equation used to calculate the average start emissions after I/M is very similar
in form to Equation 12a used to calculate the average running emissions after I/M. Several
of the parameters are the same such as the fraction of high emitters in the fleet, the waiver
rate, the waiver repair percentage, and the non-compliance rate. The principal differences
are the different IDR rates (the start IDRs are calculated in Equations 32 and 33), and the
different after repair emission levels. Equation 34 is used to calculate the After I/M start
emissions (S_EEVI). S_IDR is the start emission IDR from Equations 32 and 33, and
S_RLEV is the after successful repair emission level (in units of grams). The variable
S_RLEV is used in place of the variables N*R (normal emission level times the after repair
emission level percentage) used in the running emissions calculation.
Equation 34 is used to calculate the average emissions of the fleet after I/M, and is
used in the "sawtooth" methodology for I/M start emissions.
M6IM001.WPD DRAFT 54 Mar 24, 1999
-------
DRAFT
S_EIM = N*(1-X) + H*X*(1-S_IDR) + X*S_IDR*W*H*RW +
S_RLEV*X*S_IDR*FIX + H*x*s_iDR*NC Eqn 34
5.6 I/M "Sawtooth"
The I/M credits for start emissions will also utilize the 'sawtooth' algorithm in the
final calculation steps. This algorithm is virtually identical in structure to the ones
presented in Section 4 for the running emissions. The structures used to model change of
ownership and RSD are the same. Because the structure is the same, the methodology will
not be repeated in this section. The only difference between the start and running
algorithms are the actual emission rate parameters and values which are described in the
previous sections. These include the normal and high emission levels, the IDRs, and the
repair effects.
5.7 Remote Sensing and High Emitter Profile Start Emissions Parameters
Currently, the same remote sensing and high emitter profile parameters will be used
for the start emissions as were used for the running emissions. In the case of RSD this may
introduce some error since RSD is defined to be a warm emission test, and is not designed
to identify high start emitters or screen out low start emitters. Presumably a high emitter
profile which correctly profiles high and low start emissions can also be developed.
However, it is likely to differ from the one used for running emissions.
M6IMOO1.WPD DRAFT 5 5 Mar 24, 1999
-------
DRAFT
6.0 I/M Credits for Non-IM240 Tests
The previous sections discussed the general algorithm and methodology used to
develop the I/M credits for MOBILE6. The EVI240 test was used as the basis for the credits
because of the large amount of EVI240 data which are available to develop the IDR
estimates and the after repair levels. I/M credits for other tests are also needed such as the
Idle test, the 2500 RPM / Idle test, and the ASM tests. The algorithm used to
mathematically implement these test types in MOBILE6 is analogous to the EVI240
algorithm. The difference between the various I/M test types in MOBILE6 will be based
on the differences in the IDRs for each test.
6.1 Other I/M Tests
The MOBILE6 model will also compute I/M credits for tests other than the EVI240
test. The test options which will be built into the model are (1) Idle test, (2) 2500 RPM
/ Idle test, (3) ASM tests, and (4) On-board Diagnostic (OBD) I/M tests. In addition,
MOBILE6 will have the flexibility to model user defined test(s), or future test(s) which are
currently unspecified.
The default I/M tests in addition to the EVI240 test which MOBILE6 will able to
model are:
1. Annual Two-Mode ASM 2525/5015 with Phase-in Outpoints
2. Annual Two-Mode ASM 2525/5015 with Final Outpoints
3. Annual Single-Mode ASM 5015 with Phase-in Cutpoints
4. Annual Single-Mode ASM 5015 with Final Cutpoints
5. Annual Single-Mode ASM 2525 with Phase-in Cutpoints
6. Annual Single-Mode ASM 2525 with Final Cutpoints
7. Annual Idle Test
8. Annual 2500 RPM / Idle Test
9. Biennial Two-Mode ASM 2525/5015 with Phase-in Cutpoints
10. Biennial Two-Mode ASM 2525/5015 with Final Cutpoints
11. Biennial Single-Mode ASM 5015 with Phase-in Cutpoints
12. Biennial Single-Mode ASM 5015 with Final Cutpoints
13. Biennial Single-Mode ASM 2525 with Phase-in Cutpoints
14. Biennial Single-Mode ASM 2525 with Final Cutpoints
15. Biennial Idle Test
16. Biennial 2500 RPM / Idle Test
M6IM001.WPD DRAFT 56 Mar 24, 1999
-------
DRAFT
6.2 ASM Tests
Unfortunately, new paired ASM and FTP test data are not available on any ASM
I/M tests in-order to compute new and specific IDR rates or repair effectiveness rates. As
a result, the relative size of the I/M credits of these tests versus the EVI240 will remain the
same between MOBILES and MOBILE6. This was accomplished by first computing the
ratio of the MOBILES I/M credit value for an alternative ASM test over the MOBILES I/M
credit value for the EVI240 at final cutpoints of 0.8 HC / 15 CO / 2.0 NOx. When done for
each combination of model year, age and pollutant, this produces a large array of ratios (25
ages x 18 model year x 3 pollutants). Rather than store all those ratios in the MOBILE6
program, the ratio data were reduced by fitting it to a linear-quadratic equations using least
squares regression. The independent variables in the regression were age and model year.
The age range is from 1 to 25 and the model year range is from 81 through 98. The 98
model year credits will be used to represent all subsequent model years. The equation form
is:
ASM = A * age + B * ageA2 + C * model year + D Eqn 35
Separate equation coefficients (A, B, C and D) were developed for each ASM test,
cutpoint group, and pollutant. They are shown in Tables 7a and 7b below. Table 7a
provides the coefficients for the Final ASM cutpoints and Table 7b shows the Phase-in
ASM coefficients. Within each of these tables different coefficients were also developed
for vehicle ages which are less than or equal to 10 years, and greater than 10 years. These
ratios are then multiplied by the MOBILE6 EVI240 IDR at the 0.80 HC / 15 CO / 2.0 NOx
cutpoints to compute the MOBILE6 ASM DDR. This is done for both running and start
IDRs. After computation, the ASM IDR is used in Equation 12a to compute the ASM
After I/M line, and the I/M credits. Typically, the ASM ratioes which are applied to the
EVI240 credits are in the range of 0.60 to 1.30. This may lower or boost the EVI240 credits
by 0.30 times or raise by 0.40 times. The lower ratios prevail for HC and CO emissions,
and the higher ratios are occasionally seen for NOx emissions at the lower ages. Also, the
ratios are typically very similar to each other within a given ASM test type and pollutant
- generally ranging from 0 to 10 percent different within a model year group.
The advantage of this approach is that it enables the ASM I/M test procedure credits
to be easily assimilated into the MOBILE6 I/M approach. It also preserves a similar
relative effectiveness of ASM versus EVI240 as was present in the MOBILES model. This
is reasonable since no new ASM data are available in conjunction with FTP data to update
the ASM credits. One drawback of this approach is that it does not update the effect of
different after repair levels, and assumes that the ASM after repair levels are the same as
those for the EVI240. This means that the after repair levels for the 0.8/15/2.0 HC, CO and
M6IM001.WPD DRAFT 57 Mar 24, 1999
-------
DRAFT
NOx IM240 outpoints will be used for the final ASM outpoint after repair levels. Similarly,
the 1.2/20/3.0 HC, CO and NOx IM240 outpoints will be used for the phase-in ASM
outpoint after repair levels. Also, it assumes that the ratio between the ASM and EVI240
credits in MOBILES based on FTP emissions can be equally applied for both running and
start ASM credits in MOBILE6.
M6IMOO1.WPD DRAFT 5 8 Mar 24, 1999
-------
DRAFT
Table 7a
ASM I/M IDR Coefficients for Final ASM Cutnoints
Description
Test
ASM 2525
ASM 2525
ASM 2525
ASM 5015
ASM 5015
ASM 5015
ASM 2mod
ASM 2mod
ASM 2mod
Cutpoint
Phase -in
Phase -in
Phase-in
Phase-in
Phase-in
Phase -in
Phase-in
Phase-in
Phase -in
Pollutant
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
Test
ASM 2525
ASM 2525
ASM 2525
ASM 5015
ASM 5015
ASM 5015
ASM 2mod
ASM 2mod
ASM 2mod
Cutpoint
Phase-in
Phase-in
Phase -in
Phase -in
Phase-in
Phase-in
Phase -in
Phase-in
Phase-in
Pollutant
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
For AGE <= 10
CoeffA
0.001759
0.007035
-0.05733
-0.00494
0.003682
-0.12004
-0.006005
9.809e-04
-0.1461
CoeffB
1.588e-04
-1.826e-04
0.002875
5.766e-04
5.1103e-05
0.006165
6.0291e-04
1.5345e-04
0.007589
CoeffC
-0.001383
0.001893
-0.03234
-0.002577
0.001625
-0.02997
-0.001904
0.002478
-0.036311
CoeffD
0.9655
0.6641
4.1179
1.0894
0.6787
4.1908
1.0791
0.6573
4.9515
For AGE > 10
CoeffA
-0.005209
-0.002926
-0.001853
-0.005936
-0.003783
-0.004784
-0.004063
-0.002706
-0.005176
CoeffB
1.161e-04
6.517e-05
3.9017e-05
1.3247e-04
8.5342e-05
1.1196e-04
9.1144e-05
6.0543e-05
1.1762e-04
CoeffC
-0.001458
0.001498
-0.006412
-0.002217
0.001490
5.749e-04
-0.001374
0.002150
-0.002785
CoeffD
1.0459
0.7751
1.5271
1.1102
0.7627
0.9120
1.0614
0.7326
1.2899
M6IM001.WPD DRAFT
59
Mar 24, 1999
-------
DRAFT
Table 7b
ASM I/M IDR Coefficients for Phase-in ASM Cutnoints
Description
Test
ASM 2525
ASM 2525
ASM 2525
ASM 5015
ASM 5015
ASM 5015
ASM 2mod
ASM 2mod
ASM 2mod
Cutpoint
Phase-in
Phase-in
Phase -in
Phase -in
Phase-in
Phase-in
Phase -in
Phase -in
Phase-in
Pollutant
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
Test
ASM 2525
ASM 2525
ASM 2525
ASM 5015
ASM 5015
ASM 5015
ASM 2mod
ASM 2mod
ASM 2mod
Cutpoint
Phase -in
Phase -in
Phase-in
Phase-in
Phase-in
Phase -in
Phase-in
Phase-in
Phase -in
Pollutant
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
For AGE <= 10
CoeffA
-0.0301
0.00171
0.0289
-0.03507
1.775e-05
-0.07537
-0.03397
-3.874e-04
-0.3024
CoeffB
0.002811
6.151e-04
-0.001775
0.003147
7.131e-04
0.003535
0.003077
7.258e-04
0.01462
CoeffC
-9.764e-04
0.00390
0.015844
0.002789
0.006458
5.805e-04
-2.039e-04
0.004660
-0.10688
CoeffD
0.7324
0.2676
-1.0368
0.4213
0.05215
0.9142
0.6986
0.2228
11.890
For AGE > 10
CoeffA
-0.01390
-0.00747
-0.00118
-0.01281
-0.007068
-0.005603
-0.01242
-0.00714
-0.01342
CoeffB
3.1742e-04
1.698e-04
7.0546e-05
2.945e-04
1.6312e-04
1.7994e-04
2.8523e-04
1.6475e-04
3.5207e-04
CoeffC
-0.001254
0.00331
0.00833
0.003960
0.005987
0.01033
6.932e-04
0.004418
-0.02177
CoeffD
0.8387
0.4557
-0.2571
0.3707
0.2188
-0.3290
0.6740
0.3669
2.8099
M6IM001.WPD DRAFT
60
Mar 24, 1999
-------
DRAFT
6.3 Idle and 2500RPM/Idle Tests
The I/M credits for the Idle and 2500RPM/Idle tests were not developed like the
ASM credits by ratioing the MOBILES Idle test results with the MOBILES IM240 results
and applying the ratio to the MOBILE6IM240 results to get the MOBILE6 Idle test credits.
Although, they could have been developed this way. Instead, the Idle and Idle/2500 RPM
test credits were developed from a new analysis of the available paired Idle / 2500RPM/Idle
and FTP data sources collected by EPA from 1981 through 1998.
6.3.1 Available Data
Two primary EPA datasets were available. The first dataset is called the "4MID"
dataset. The abbreviation "4MID" stands for "Four Mode Idle dataset". It contains
virtually all of EPA's paired Idle and FTP data collected at EPA's various labs from 1981
through 1998. The four mode test is a special EPA Idle I/M test procedure developed for
research work that simulates in-use Idle tests. The first mode is an unpreconditioned idle,
the second mode is a 2500 RPM segment used to precondition the third Idle mode, and
used to pass or fail vehicles for the 2500RPM/Idle test. The third mode is a preconditioned
Idle, and the fourth mode is an idle in drive mode. Only the 2500 RPM mode and the third
mode (pre-conditioned Idle) were used to develop the credits. Only the HC emissions from
the 2500 RPR mode were used in the development of the 2500RPM/Idle credits. The
analogous CO 2500 RPM mode readings were not used because of their tendency to
produce false failures due to evaporative canister purge during the 2500 RPM mode. The
preconditioned Idle test was used in both the Idle test and the 2500RPM/Idle test credits.
The unpreconditioned Idle mode and the Idle in Drive modes were not used for the I/M
credit development.
Test results from the Restart /Idle test used to test some early 1980's Ford vehicles
were not used in this analysis due to their inconsistent availability in the dataset. The effect
of this is thought to be very negligible. However, since the basis of the IDR consists only
of High emitting vehicles, use of the Four mode test instead of the Restart / Idle test for
Ford vehicles could potentially overstate the Idle test credits slightly if the higher readings
from the Four Mode test identify more high emitters that the Restart / Idle test would
identify.
The second primary dataset was the "IMLane" dataset. It consisted of I/M lane Idle
and 2500RPM/Idle test results from EPA's pilot I/M lane test program conducted in both
Hammond, IN and Phoenix, AR by ATL. These data were paired with vehicle FTP data
collected at ATL's laboratory. The test procedure consisted of a 2500RPM mode, and a
subsequent preconditioned Idle mode. The unpreconditioned Idle and the Idle in Drive
modes were not performed. The advantage of these data over the 4MID sample is that they
were collected in an actual I/M lane rather than in the EPA laboratory like the 4MID
M6IMOO1.WPD DRAFT 61 Mar 24, 1999
-------
DRAFT
sample.
For the final results, both databases were combined together to produce overall IDR
rates for the Idle test and the 2500RPM/Idle test. Despite the slight differences in the I/M
test procedures, the combination of the data makes sense for several reasons. First, it
produces a larger sample of vehicles. This is important because for this analysis only the
High emitters are used to compute the IDRs, and the number of High emitters can get small
in some model year groups. Also, both databases seem to complement each other in terms
of model year coverage. For example, the "4MID" sample has a large preponderance of its
data in the 1981 and 1982 model years; however, it does have some newer mid 1990's
vehicles and trucks. The ATL sample on the other hand contains only cars, and is mostly
represented by late 1980's to early 1990's cars. Tables 8a and 8b show the model year and
technology breakdown for both databases.
M6IM001.WPD DRAFT 62 Mar 24, 1999
-------
DRAFT
MY
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
Table 8a
Four Mode Idle / 2500RPM Idle and FTP Test Pairs
Cars
CARS
962
125
87
32
90
41
16
15
22
TBI
15
66
122
44
52
52
64
60
35
46
4
2
4
PFI
29
5
59
34
61
86
92
103
82
85
59
37
16
27
2
Trucks
CARS
120
45
10
48
63
17
TBI
13
23
1
PFI
4
1
6
41
2
2
1
M6IM001.WPD DRAFT
63
Mar 24, 1999
-------
DRAFT
MY
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
Table 8b
IM Lane Idle / 2500RPM Idle and FTP Test Pairs
Idle Test
CARS
39
37
22
21
14
11
9
4
1
1
TBI
1
3
18
56
65
61
39
41
34
25
6
2
PFI
2
1
11
29
48
47
48
61
53
33
17
18
6
2500 RPM / Idle Test
CARS
39
37
22
21
14
11
9
4
1
1
TBI
1
3
18
56
63
61
39
40
34
25
5
2
PFI
2
1
10
29
47
47
48
60
53
33
17
18
6
6.3.2 Idle and 2500RPM/Idle Test IDRs
The calculation of the IDRs for the Idle and 2500RPM/Idle tests is very similar to
the calculation done for IM240 IDRs in Section 3.9. One difference is that IDRs for a range
of cutpoints was not performed. Instead only one set of Idle and 2500RPM/Idle cutpoints
were developed. These were at the CO/HC cutpoints of 1.2%CO and 220ppm HC. Also,
IDRs for only HC and CO emissions for running and start were developed. Idle and
2500RPM/Idle IDRs for NOx emissions were not developed. Neither the Idle Test or the
2500RPM/Idle test will produce NOx benefits or NOx "Dis-benefits" for MOBILE6. In
comparison, MOBILES contained NOx "Dis-benefits" if an Idle or 2500RPM Idle test were
performed.
M6IM001.WPD DRAFT
64
Mar 24, 1999
-------
DRAFT
Test
Idle
2500/Idle
Test
Idle
2500/Idle
Table 9a
Idle and 2500RPM / Idle Test IDRs for Each Sample
IDRs Based on I/M Lane Sample
Hot Running LA4 HC
Carb
63.3
76.5
PFI
58.7
59.3
TBI
53.2
53.9
Cold Start HC
Carb
41.9
48.6
PFI
39.1
40.2
TBI
33.9
34.8
Hot Running LA4 CO
Carb
54.9
68.8
PFI
57.5
57.5
TBI
60.6
60.6
Cold Start CO
Carb
29.1
29.1
PFI
23.6
23.6
TBI
20.9
20.9
Test
Idle
2500/Idle
Test
Idle
2500/Idle
IDRs Based on Four Mode Sample
Hot Running LA4 HC
Carb
48.8
66.1
PFI
74.3
74.3
TBI
52.2
61.6
Cold Start HC
Carb
20.2
24.4
PFI
42.6
42.6
TBI
17.7
25.4
Hot Running LA4 CO
Carb
53.4
63.8
PFI
81.1
81.1
TBI
40.7
55.7
Cold Start CO
Carb
21.4
27.1
PFI
57.8
57.8
TBI
30.1
33.9
Table 9a shows the Hot Running LA4 and Cold Start DDR rates for the Idle and
2500RPM/Idle tests for each of the two datasets. It is further broken down into three
technology groups. These are Carbureted, Throttle Body Injection (TBI), and Ported Fuel
Injection (PFI). The IDRs were not made a function of model year because of the small
sample sizes in many individual model years. Table 9b shows the DDR results for the
combined dataset. The two datasets were combined together based on total emissions from
M6IM001.WPD DRAFT
65
Mar 24, 1999
-------
DRAFT
the high emitters rather than on the number of vehicles in the sample. The IDRs are shown
as a percentage in both tables, but will be programmed into MOBILE6 as fractions. They
represent the fraction of emissions from high emitters which are identified by the
prospective I/M test. Separate IDRs for each pollutant and technology were developed for
Hot Running LA4 emissions and Start emissions based on Bagged FTP data.
Table 9b
Idle and 2500RPM / Idle Test IDRs Based on the COMBINED Sample
Test
Idle
2500/Idle
Test
Idle
2500/Idle
IDRs Based on I/M Lane Sample
Hot Running LA4 HC
Carb
54.6
70.2
PFI
63.5
63.9
TBI
52.8
56.8
Cold Start HC
Carb
25.5
30.3
PFI
40.8
41.3
TBI
29.5
32.3
Hot Running LA4 CO
Carb
54.0
65.9
PFI
63.0
62.9
TBI
53.5
58.8
Cold Start CO
Carb
23.3
27.6
PFI
37.8
37.8
TBI
25.1
26.8
6.3.3 After Repair Emission Level for Idle and Idle/2500 Tests
The Idle Test after repair emission levels for MOBILE6 were calculated from a
dataset which was used for MOBILES development. It consisted of 36, 1981 and later
vehicles which initially failed the idle test, were repaired, and passed the final idle test at
standard cutpoints. These data were collected as part of an EPA test program conducted
to evaluate the effect of repair on idle test failures. The repairs were conducted by qualified
technicians. The vehicle sample mean FTP emission values after Idle test I/M repair were
found to be 1.89 g/mi HC and 19.49 g/mi CO. These compare with means of 1.26 g/mi
HC and 13.46 g/mi CO for the EVI240 at the 1.2/20 HC and CO cutpoint. Idle test repair
effects for NOx emissions are not computed because MOBILE6 will not give NOx benefits
or disbenefits to an idle test program.
The ratio of the idle test after repair FTP emission level to the EVI240 after repair
FTP emission level at 1.2/20/3.0 cutpoints is computed from the data and used to generate
M6IM001.WPD DRAFT
66
Mar 24, 1999
-------
DRAFT
the after repair idle test emission level for running LA4 emissions. A consistent ratio based
on the FTP will be used for all mileages, vehicle types, and model years. The ratios which
are used for HC and CO are:
HC Ratio: 1.89 g/mi / 1.26 g/mi = 1.50
CO Ratio: 19.49 g/mi / 13.46 g/mi = 1.45
They are used in MOBILE6 to generate the idle test after repair running LA4
emission level by multiplying the ratio by the EVI240 after repair emission level at
1.2/20/3.0 cutpoints. The same after repair emission levels will be used for the Idle test and
the Idle/2500 RPM test.
6.4 OBD I/M Tests
This document does not explicitly cover vehicles which are equipped with an OBD
system. However, most OBD equipped vehicles will continue to receive exhaust based I/M
tests such as the EVI240 or the Idle test for much of their early lives. Thus, the topic is
mentioned briefly in this document as an introduction. For more complete details on EPA's
modeling of OBD equipped vehicles (1996+ model years) please read EPA document
M6.EXH.007 "Determination of Emissions, OBD, and I/M Effects for Tierl, TLEV, LEV,
and ULEV Vehicles".
The OBD system is an electronic diagnostic system built into most 1996 and later
and some 1994 and 1995 model year vehicles. It is designed to (1) continuously monitor
the performance of the car's emission control system, and detect serious problem(s) which
cause the vehicle's FTP emissions to exceed 1.5 times its applicable certification standards,
(2) register a code in the vehicle's computer and turn on a dashboard warning light to notify
the owner. The system will also have the capability to be electronically accessed in an I/M
lane. The vehicle will be required to pass the OBD test (no trouble codes are present) in-
order to pass the state I/M program requirements.
In MOBILE6 an I/M program conducting an OBD check on properly equipped OBD
vehicles will be assigned an DDR of 90 percent (fraction 0.90). This value will be given
regardless of whether an exhaust I/M test such as the EVI240 or the ASM test is performed
or not performed. Also, the with and without technician training levels in an OBD I/M
program will be equivalent. It is assumed that the technicians specializing in OBD
diagnosis and repair will either be fully qualified, or not involved in the industry.
M6IM001.WPD DRAFT 67 Mar 24, 1999
-------
DRAFT
APPENDIX A
Running LA4 Emissions from 1990-93 MY PFI Normal Emitters
Figure A-1
HC Emissions from Normals
20
MILEAGE
40
60
80
100
120
140
160
180
200
Figure A-2
CO Emissions from Normals
20.
18.
16.
14.
12.
10.
8.
6.
4.
0 20 40 60 80 100 120 140
MILEAGE
160
180
200
M6IM001.WPD DRAFT
68
Mar 24, 1999
-------
DRAFT
Figure A-3
NOx Emissions from Normals
20
MILEAGE
40
60
80
100
120
140
160 180
200
M6IM001.WPD DRAFT
69
Mar 24, 1999
-------
DRAFT
APPENDIX B
Sample Calculations
Sample calculation for determining percentage of HC running emission Highs in the fleet at age = 5 for 1990-
1993 PFI technology.
Calculating Line A in Figure 1 (basic emission rate)
X< 21.27 A = 0.0508 +0.0013 * mile
X > 21.27 A = 0.0508 + 0.0013 * mile + (mile - 21.27) * 0.0023
from Table 3, mile = 49.835
X is the inflection point of the basic emission rate (thousand mile units).
See the document "Determination of Running Emissions as a Function of Mileage for 1981-1993 Model
Year Light-Duty Vehicles."
A = 0.181g/miHC
Calculating Line B in Figure 1 (normal emitter rate)
B = 0.0249+ 0.00113* mile
B = 0.081g/miHC
Calculating Line C in Figure 1 (high emitter rate)
C= 1.367 g/miHC
Calculating Line D in Figure 1 (After I/M repair emission rate)
D = ( 2.24 - 0.07595 * Age ) * B
0=1.86*0.081=0.151
Calculating percentage of Highs from equation 7.
Highs = (A - B) / © - B)
Highs = 0.078 = (0.181 - 0.081) / (1.367 - 0.081)
% Highs = 7.8 percent
M6IM001.WPD DRAFT 70 Mar 24, 1999
-------
DRAFT
APPENDIX C
Periodic I/M and RSD / Change of Ownership Sawtooth Illustrations
Figure 1 - Annual I/M with RSD
SSI
Uj
O
0
^CMf/J^lA
c«Mf>/r^ 5e<5M£(JrZ8
•ffSMo/Tie / /-'-f.sMEAjr^ 8
'/Z
M6IM001.WPD DRAFT
71
Mar 24, 1999
-------
DRAFT
Figure 2 - Biennial I/M with RSD
2 i 4 5 & 1 9 9 10 il II 13 14-
YEAfiS
4
M6IM001.WPD DRAFT
72
Mar 24, 1999
-------
DRAFT
Appendix D
Description of the FORTRAN Algorithm Adopted from EPA Document
M6.IM.001
Used to Code the I/M Methodology in MOBILE6
TASK:2-663 DRAFT:
12/01/99
DynTel Report:
Mobile 6 IM Benefits Methodology for 1981 through 1993 Model Year Light
Vehicles
Employee: Robert Ducharme
1. Introduction
In most inspection and maintenance programs vehicles are inspected annually or
biennially. However, some inspections are prompted by special events such as change of
ownership (COEVI) or identification of high emitting vehicles using a remote sensing device
(RSD). The objective of this report is to derive an equation for the emissions from a fleet
of vehicles that is subject to both periodic and selective (RSD+COEVI) EVI programs. The
most general form of this equation allows for an arbitrary period N between inspections and
an arbitrary grace period GPRD before each vehicle receives its first test. However, both
N and GPRD must be an integer number of years.
For the purposes of modeling, light duty gasoline vehicles and trucks are classed
either as normal or high emitters. High emitters are the vehicles with broken emission
control systems. The influence of inspection and maintenance programs is to reduce the
basic exhaust emission levels from high emitting vehicles compared to what they would be
if no EVI program were in force. No EVI correction is required for normal emitting vehicles.
This report is based on the inspection and maintenance methodology described in
the US EPA draft report M6.EVI.001 though there are some differences. One such point of
departure is that the EVI sawtooth methodology from Mobile 5 is replaced using an
algebraic approach for calculating the benefits of EVI programs that does not require the use
of sawtooth diagrams. This has led to two refinements of the EPA model. Firstly, the basic
emission factor lines drawn straight in the sawtooth diagrams are slightly curved in reality.
M6EVI001.WPD DRAFT 73 Mar 24, 1999
-------
DRAFT
This curvature has now been taken into account. Secondly the sawtooth method applies the
benefits of periodic IM testing on a single day (October 1) during each relevant twelve
month test period preceding the emissions evaluation date. Here, a continuous model is
used that assumes vehicles are always tested on the anniversaries of their sales. It is further
assumed that new vehicle sales are uniform throughout each model year so that the
distribution of the anniversaries of those sales is also uniform in future model years. The
notation used in this report is not month specific so that the mathematical formulation is
equally applicable for both January 1 and July 1 calculations.
2. IM240 tests
EPA have worked out an explicit equation for the quantity that must be subtracted from
the basic emission factor of any high emitting 1981-1993 model year light duty gasoline
vehicles and trucks in order to take into account the benefits of having an JM240 program
in force. There is no IM correction for normal emitting vehicles. The form of this correction
factor is readily deduced from eqn 12a and eqn 34 (ref M6.IM.001) to be:
IMCF(JDX,AIM) = XIM(JDX,AIM)* IDR
*[HIM(JDX, AIM) * (W * R W + NC - 1) + A * FIX ] (1)
where the symbols have the following meaning
AIM: Integer age of a vehicle in years on the date of its IM test previous to the
emissions evaluation date. It is assumed that vehicles are always tested
on their anniversaries of their sales.
JDX: Integer model year index of a vehicle referred to the year ICY and month
MEVAL of the emission factor calculations.
XIM: Fraction of the fleet composed of high emitters on the date of the IM test.
HIM: High average emission factor on the date of the IM test.
DDR: Fraction of all the high emitters in a target group identified by an IM test.
W: Fraction of all the identified high emitters that get a repair cost waiver.
RW: Fraction of the high emitter level that waived vehicles are repaired after
IM.
NC: Fraction of identified high emitters which are in non-compliance of the
IM program.
FIX: Fraction of identified high emitters which get repaired to pass the test.
A: Average emission level from vehicles after they have been repaired and
passed an IM test.
M6IM001.WPD DRAFT 74 Mar 24, 1999
-------
DRAFT
Eqn (1) normally has a negative value and applies for both running (ISR=1) and start
(ISR=2) emissions. However, the calculation of the after repair level A and the
identification rate IDR is different depending on the value of ISR. The details of these
calculations can be found in M6.EVI.001 together with default values for R, RW and NC.
The value of FIX is 1-W-NC. The calculation of XEVI and HIM are discussed in section 4.
3. Other IM Tests Types
Eq. (1) is valid for the following additional EVI test types.
1. Two-Mode ASM 2525/5015 with phase-in cutpoints
2. Two-Mode ASM 2525/5015 with final cutpoints
3. Single-Mode ASM 5015 with phase-in cutpoints
4. Single-Mode ASM 5015 with final cutpoints
5. Single-Mode ASM 2525 with phase-in cutpoints
6. Single-Mode ASM 2525 with final cutpoints
7. Idle test
8. 2500 RPM/Idle test
The test type affects the benefit through the high emitter identification rate IDR and the
after repairs emission level A. IDR is also corrected for RSD clean screening and high
emitter profiling. The maximum allowable value of IDR is 0.9. The after repairs emission
level A includes a correction for technician training. It cannot be higher than the high
emission factor or lower than the normal emission factor.
4. IM Emission Factors
The EPA EVI methodology for periodic (annual, biennial, triennial etc.) EVI programs
makes two simplifying assumptions.
17. All vehicles are tested on the anniversaries of their sales.
18. Vehicle sales are uniform throughout each model year.
The assumption that EVI tests always tested on the anniversaries of their sales will have to
be revised when COEVI and RSD prompted testing is considered later.
The normal (INH=1) and high (INH=2) basic emission factors for a vehicle of
model year MY on the date of an EVI test is
BIM(INH,MY,AM) =ZML(INH,MY) +KIM(AIM) *DR(INH,MY) (2)
where ZML is the zero mile level, DR is the deterioration rate, KEVI is the vehicle miles
M6EVI001.WPD DRAFT 75 Mar 24, 1999
-------
DRAFT
traveled in units of thousands of miles for a vehicle of age AIM. Note, MY=ICY-JDX+1
where JDX has previously been defined as the model year index referred to the evaluation
year ICY. Default values for ZML and DR are stored in the mobile model. The value of
KIM for AEVI>0 is readily calculated from the expression
(3)
where AMAR(I) is the annual mileage accrual rate also stored in Mobile. It is a reasonable
approximation to assume the vehicle has traveled zero miles when its first owner aquires
it so KIM(0)=0.0.
The HIM variable defined in section 2 is:
HIM =BIM(INH=2,MY,AM)
The default value of the deterioration rate for highs emitters in Mobile is zero but the user
can override this default.
The probability that a vehicle of age AIM will be a high emitter is XIM(MY, AGE).
This quantity like BEVI can also be expressed exclusively as a function of MOBILE 6
regression coefficients and KIM.
5. The NO IM case
Mobile 6 calculates emissions on January 1 or July 1. The existing method of
evaluating the uncorrected (FTP) basic emission factors on these dates is through the
equation
Normals: NOIM(INH=1,MY,JDX)=BEV(INH=1,MY,JDX) *(1-XEV(MY,JDX))
Highs: NOIM(INH=2,MY,JDX) =BEV(INH=2,MY,JDX) *XEV(MY,JDX)
(4)
where
BEV(INH,MY,JDX) =ZML(INH,MY)+KMILES(JDX)*DR(INH,MY) (5)
and XEV is the fraction of high emitters in the fleet on the evaluation date. Expression (5)
is identical to (2) except the model year index is now calculated from the evaluation date
and KMTLES has replaced KIM where KMTLES is the average vehicle mileage on the
evaluation date. Similar arguments apply to XEV and its XEVI counterpart. The method of
M6IM001.WPD DRAFT 76 Mar 24, 1999
-------
DRAFT
adjusting vehicle mileage for the month of evaluation is described in AP42 for the special
case of January 1 emissions. This method has since been extended to treat the July 1 case
FRN
JDX=4
FRC
FRN
JDX=3
FRC
FRN
JDX=2
FRC 1
JDX=lJ
Evaluation
date.
Figure 10 Shows the partitioning of each model year into segments.
as well. The general formula for KMTLES expressed in terms of KJJVI is
KMILES(l) = -FRC * KIM (I)
KMILES(JDX) = KIM(JDX)+-FRC2 * [KIM(JDX + 1) - KIM(JDX)]
--FRN2 * [KIM(JDX)-KIM(JDX -1)] (6)
where FRC is the elapsed fraction of a year since the model year changed on October 1 and
FRN=1-FRC. The fraction FRC is readily calculated in terms of the month of evaluation
MEVAL using the algorithm
DIFF = MEVAL -10
FRC(DIFF >0) = DIFF 112
FRC(DIFF < 0) = 1 - DIFF 112
M6JM001.WPD DRAFT
77
Mar 24, 1999
-------
DRAFT
Thus, FRC=0.25 on January 1 and FRC=0.75 on July 1.
6.0 Annual I/M Inspection
In annual JM programs vehicles are inspected every year on the anniversary of the sale to
their first owner. For generality, it will be assumed that vehicles do not receive their first
inspection until they have been in operation for GRPD years. Consequently, all vehicles
with model year index greater than GRPD+1 should receive one inspection in the twelve
month period preceding the date when the emissions are to be evaluated.
An essential concept in analyzing periodic JJVI programs is the fact that the age of a vehicle
AIM on the date of its previous EVI test is not a unique function of its model year index.
However, it is possible to partition each model year into two segments in such a manner
that a unique value of AIM can be assigned to each segment. This breakdown is done next
for the case of an annual EVI inspection program with a grace period of 1 year. The value
of AIM can be determined for each model year and model year segment with the help of
figure 1. For example, if emissions are to be evaluated in 1990 from a 1988 model year
vehicle then JDX=3 will be the model year index of the vehicle. The JDX=3 model year
can be divided, as can any other year, into FRC and FRN segments. Therefore, suppose that
the vehicle was purchased new in the FRC model year segment. The choice of JDX=3 and
the FRC segment give the starting point in the diagram. It is then simply a question of
counting forward an integer number of years (AIM) until the date of the previous EVI test
before the evaluation date is found. The result is AIM=2. Table 1 shows the value of AEVI
for other values of JDX.
JDX SEGMENT AIM
1 FRC 0
2 FRN 0
2 FRC 1
3 FRN 1
3 FRC 2
4 FRN 2
4 FRC 3
JDX FRN JDX-2
JDX FRN JDX-1
Table 1. Vehicle ages in annual EVI programs.
M6EVT001.WPD DRAFT 78 Mar 24, 1999
-------
DRAFT
The above table gives sufficient information to calculate periodic EVI corrections for any
model year. For example, for JDX=3 it can be seen that the correction is
PIMCF(JDX=3)=FRN*IMCF(JDX=3,AIM=1)+FRC*IMCF(JDX=3,AIM=2)
(7)
In completion of the annual EVI program problem it is necessary to treat arbitrary
values of the grace period. The JDX
-------
DRAFT
2. Move forward an integer number of years equal to the grace period.
3. If you have moved beyond the emissions evaluation date set AIM1=0.
4. Else move move forward in steps of N years until the date of the
previous EVI test. In this case, AEVI1 is equal to the total number of years
between the purchase date and previous EVI test date for the vehicle.
The algorithm for calculating AEVI2 is identical except that step 1 begins in the FRC
segment of the JDXth model year. The complication of vehicles that are too young to have
received their first test is readily handled using the convention EVICF(JDX, AEVI=0)=0. One
simple test of this algorithm is to reproduce the results in section 6 for an annual EVI
program. It is also of interest to calculate the vehicle ages for biennial programs with grace
period of 1 and 2 years. These results are given in table 2.
JDX SEGMENT AIM
1 FRC
2 FRN
2 FRC
3 FRN
3 FRC
4 FRN
4 FRC
5 FRN
5 FRC
6 FRN
6 FRC
Table 2. Vehicle ages in biennial EVI programs.
8. Selective IM programs
In periodic EVI programs all vehicles are tested every N years following an initial grace
period GPRD years after they were first bought into the fleet. In selective EVI programs,
vehicles are only tested if they meet certain criteria such as a recent change of ownership
(COEVI) or detection as a high emitter using a remote sensing device (RSD). Selective EVI
testing is usually done in areas where a periodic program is also in operation. One
important difference between selective and periodic programs is that a vehicle can be
tested at any time during the year rather than just on the anniversary of its purchase.
However, for modeling purposes it will be assumed that selective EVI tests only affect
vehicles of integer and half-integer ages.
M6EVI001.WPD DRAFT 80 Mar 24, 1999
GPRD=1
0
0
1
1
1
1
O
3
O
3
5
GPRD=2
0
0
0
0
2
2
2
2
4
4
4
-------
DRAFT
Let PSTE(JDX, AGE) and PSTL(JDX, AGE) denote the respective probabilities that
a FRC and FRN segment vehicle with model year index JDX and age AGE
(=0.5,1.0,1.5....) will have an EVI test as a result of identification in the previous six months
by either a COEVI or RSD program. These quantities can be defined more precisely as
PSTE(JDX,AGE)=FSTE(JDX,AGE) *STR(JDX,AGE) (10)
PSTL(JDX,AGE)=FSTL(JDX,AGE) *STR(JDX,AGE) (11)
where F STL and FSTE are the FRC and FRN segment fractions of all the JDX model year
vehicles eligible by virtue of having reached the age AGE for a selective EVI test and
STR(JDX,AGE)=RSD(JDX,AGE)+COIM (JDX,AGE) (12)
is the normalized probability that an eligible vehicle will be selected for a test as a result
of change of ownership or detection by a remote sensing device. For example, if
PSTE(JDX=3,AGE=1.5)=0.01
then 1% of all the JDX=3 vehicles will receive an JJVI benefit EVICF(JDX=3,AGE=1.5) as
a result of the fact they were all purchased new in the same FRC model year segment and
tested at the same age of 1.5 years.
Selective EVI tests only benefit vehicle emissions if they take place after the vehicles
previous periodic EVI test. All FRC segment vehicles
FSTE(JDX,AGE) = FRC (13)
will be eligible for selective EVI tests for integer and half integer values of AGE in the range
AIM +0.5< AGE < JDX-I (14)
providing AEVKJDX-1. This result can be deduced from figure 1. If FRC>0.5 then a
fraction
FSTE(JDX,AGE) = FRC-0.5 (15)
of the JDX model year vehicles will also be eligible for one additional test at age JDX-0.5.
Eqns (13) and (15) give the only nonzero values of FSTE. The arguments pertaining to the
FRN segment vehicles are similar. In particular, all FRN segment vehicles
FSTL(JDX,AGE) = FRN (16)
M6EVIOO1.WPD DRAFT 81 Mar 24, 1999
-------
DRAFT
will be eligible for selective IM tests for integer and half integer values of AGE in the range
AIM +0.5< AGE < JDX-2 (17)
providing AEVKJDX-2. If FRN<0.5 then all FRN segment vehicles will also be eligible for
an additional test at age JDX-1.5. Otherwise if FRN>0.5 then only a fraction
FSTL(JDX,AGE)=0.5 (18)
of the JDX model year vehicles will be eligible.
The probability CPSIME(JDX,AIM) that a vehicle purchased new in the FRC
segment of the JDX model year receiving a selective EVI test between its previous periodic
EVI test date at age AIM and the emissions evaluation date at the end of the FRC segment
of the JDX=1 model year is equal to the sum over PSTE(JDX,AGE) for all the values of
AGE satisfying equations (14) and (15). This is given by
M ,
CPSIME(JDI,AIM)= ^[PSTE(JDX,AGE = AIM +0.5*M)] (19)
where ME=2*(JDX-AEVI-0.5) is the maximum number of possible selective EVI test dates.
The arguments for vehicles in the FRN model year segment are similar with all the possible
values of the AGE variable calculable from eqns. (17) and (18). This leads to the
probability
CPSIML(JDX,AIM) = ^[PSTL(JDX,AGE = AIM +0.5* M)] (20)
with ML=2*(JDX-AIM-1.5). It is a further requirement that vehicles cannot receive the
benefits from more than one EVI test. Consequently, the values of CPSEVIE and CPSIML
cannot exceed FRC and FRN respectively.
The arguments in this section up to here have been quite formal. It is therefore
instructive to once again consider an example. Consider the case of a biennial EVI program
with selective EVI testing and a grace period of one year. Let us set JDX=4 and select a
vehicle that was bought new in the FRN model year segment. Table 2 indicates that this
vehicle will have received its previous
periodic EVI test at age AIM=1 year. From eqn (20) the probability that this vehicle will be
tested as a result of the selective EVI program is
M6EVT001.WPD DRAFT 82 Mar 24, 1999
-------
DRAFT
CPSIML(JDX=4,AIM=1) =PSTL(JDX=4,AGE=1.5) + PSTL(JDX=4,AGE=2.0)
+ PSTL(JDX=4, AGE=2.5)
By contrast, the vehicles belonging to the FRC segment of the JDX=4 model year segment
are all old enough to have received their second periodic EVI test at age AEVI=3. In this case
eqn (19) gives
CPSIME(JDX=4,AM=3) = PSTE(JDX=4,AGE=3.5)
where this expression contains only one term because a relatively short period elapses
between the date when the vehicles receive their periodic EVI test and the emissions
evaluation date.
Each PSTE(JDX,AGE) term in eqn (19) gives the probability that a vehicle bought
new in the FRC segment of the JDX model year will receive a selective EVI test at age AGE.
The benefit that arises from such a test is therefore PSTE(JDX, AGE)*EVICF(JDX, AGE).
These benefits can therefore be summed over all the possible selective EVI test dates to give
the total benefit from all the selective EVI tests carried out on the FRC segment vehicles to
be
SIMCFE(JDX,AIM) =
Mg
^[PSTE(JDX,AGE = AIM + 0.5 * M) * IMCF(JDX,AGE = AIM +0.5*M)]
M=l
(21)
The total benefit from all the selective EVI tests carried out on the FRN segment vehicles
is then similarly
SIMCFL(JDX,AIM) =
M L
^[PSTL(JDX,AGE = AIM + 0.5 * M) * IMCF(JDX,AGE = AIM +0.5*M)]
(22)
Here, EVICF is evaluated using the vehicle mileage equation
M6IM001.WPD DRAFT 83 Mar 24, 1999
-------
DRAFT
j
KIM (J + 0.5) = 0.5* AMAR(J + 1) + ^AMAR(I) (23)
7=1
for all half-integer ages.
9. Calculation of IM Benefits
Suppose a selective IM program is operating alongside the annual program. The
generalization of eqn. (9) to include the effect of the selective testing is then
PMCF(JDX)=(FRN-CPSML(JDX,AM1)) *IMCF(JDX,AM1)
+ SIMCFL(JDX,AM1)
+ (FRC-CPSIME(JDX,AM2)) *IMCF(JDX,AM2)
+ SIMCFE(JDX,AIM) (24)
There are two points to note. Firstly, the earlier and later SEVICF terms are included to
account for the benefits of the selective EVI tests that take place after the periodic EVI tests.
Secondly, the CPSEVI terms are subtracted from the FRC and FRN fractions so that the
selectively tested vehicles do not also receive a benefit for their earlier periodic test.
It is instructive to evaluate eqn. (24) for the correction to July 1 emissions arising
from an annual EVI program with COEVI. Let JDX=3. Annual I/M programs are treated in
section 6 where the values AEVI1=1 and AIM2=2 can be read from table 1. With this
information, the arguments in section 8 then give the cumulative probabilities and selective
I/M correction factors for this problem to be
CPSIME(JDX=3,AM=2)=0.25 *STR(JDX=3,AGE=2.5)
CPSIML(JDX=3,AM=1)=0.25*STR(JDX=3,AGE=1.5)
SIMCFE(JDX=3,AM=2)=CPSIME(JDX=3,AM=2)* MCF(JDX=3,AGE=2.5)
SIMCFL(JDX=3,AM=2)=CPSML(JDX=3,AM=1)* MCF(JDX=3,AGE=1.5)
having set FRC=0.75 and FRN=0.25. Here, the value of STR depends on the rate of
selective testing. For example, if a COEVI program is in operation in an area with 16% per
annum change of ownership then STR=0.08. In this case, eqn (24) simplifies to
PMCF(JDX=3)=0.23*IMCF(JDX=3,AGE=1)
M6EVI001.WPD DRAFT 84 Mar 24, 1999
-------
DRAFT
+ 0.02* IMCF(JDX=3,AGE=1.5)
+ 0.73*IMCF(JDX=3,AGE=2)
+ 0.02* IMCF(JDX=3,AGE=2.5)
where IMCF can be calculated directly from eqn (1) for each of the 4 vehicle ages.
M6IM001.WPD DRAFT 85 Mar 24, 1999
-------
DRAFT
APPENDIX E
Statistical Diagnostics for Running Emissions IDR Determination
-> REGRESSION
-> /DESCRIPTIVES MEAN STDDEV CORR SIG N
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hcrun_id
-> /METHOD=ENTER ln_hccut ln_cocut
**** MULTIPLE REGRESSION ****
Equation Number 1 Dependent Variable.. HCRUN_ID HCRun ID
Descriptive Statistics are printed on Page 2
Block Number 1. Method: Enter LN_HCCUT LN_COCUT
Variable(s) Entered on Step Number
1.. LN_COCUT
2 . . LN_HCCUT
Multiple R .90947
R Square .82713
Adjusted R Square .82246
Standard Error .06411
Analysis of Variance
DF Sum of Squares Mean Square
Regression 2 1.45516 .72758
Residual 74 .30413 .00411
F = 177.03226 Signif F = .0000
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
LN_HCCUT -.136503 .010483 -.157390 -.115615 -.629362
LN_COCUT -.106888 .007869 -.122568 -.091209 -.656531
(Constant) 1.145095 .026063 1.093164 1.197027
in
Variable T Sig T
LN_HCCUT -13.021 .0000
LN_COCUT -13.583 .0000
(Constant) 43.936 .0000
-> REGRESSION
-> /DESCRIPTIVES MEAN STDDEV CORR SIG N
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT corun_id
-> /METHOD=ENTER ln_hccut ln_cocut
**** MULTIPLE REGRESSION ****
Equation Number 1 Dependent Variable.. CORUN_ID CORun ID
Block Number 1. Method: Enter LN HCCUT LN COCUT
M6IM001.WPD DRAFT 86 Mar 24, 1999
-------
Variable(s) Entered on Step Number
1.. LN_COCUT
2.. LN HCCUT
DRAFT
Multiple R .90658
R Square .82188
Adjusted R Square .81707
Standard Error .06736
Analysis of Variance
Regression
Residual
170.72789
DF Sum of Squares
2 1.54920
74 .33574
Signif F = .0000
Mean Square
.77460
.00454
Variable
Variables in the Equation
B SE B 95% Confdnce Intrvl B
LN HCCUT
LN COCUT
(Constant)
- .107306
- .129819
1.188020
.011014
.008268
.027384
- .129253
- .146293
1.133456
- .085360
- .113344
1.242584
- .477976
- .770339
Variable
LN_HCCUT
LN_COCUT
(Constant)
T Sig T
-9.742 .0000
-15.702 .0000
43.384 .0000
-> * Curve Estimation.
-> TSET NEWVAR=NONE .
-> CURVEFIT /VARIABLES=noid WITH nocut
-> /CONSTANT
-> /MODEL=CUBIC
-> /PRINT ANOVA
-> /PLOT FIT.
Dependent variable.. NOID
Listwise Deletion of Missing Data
Multiple R .99902
R Square .99805
Adjusted R Square .99658
Standard Error .01860
Analysis of Variance:
DF Sum of Squares
Regression
Residuals
.70707598
.00138343
Method.. CUBIC
Mean Square
.23569199
.00034586
F = 681.46957 Signif F = .0000
Variables in the Equation
Variable B SE B Beta T Sig T
NOCUT
NOCUT**2
NOCUT**3
(Constant)
.756842 .102036 3.175112
-.368671 .037175 -9.352562
.040631 .004083 5.358327
.545291 .082060
7.417 .0018
-9.917 .0006
9.951 .0006
6.645 .0027
M6IM001.WPD DRAFT
87
Mar 24, 1999
-------
DRAFT
APPENDIX F
Statistical Diagnostics for Start Emissions IDR Determination
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_strt_
-> /METHOD=ENTER In hccut In cocut
**** MULTIPLE REGRESSION ***
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_STRT_ HC Strt ID
Block Number 1. Method: Enter LN HCCUT LN COCUT
Variable(s) Entered on Step Number
1.. LN_COCUT
2.. LN HCCUT
Multiple R .85506
R Square .73113
Adjusted R Square .70669
Standard Error .11633
Analysis of Variance
DF Sum of Squares Mean Square
Regression 2 .80951 .40476
Residual 22 .29769 .01353
F = 29.91216 Signif F = .0000
- Variables in the Equation
B SE B 95% Confdnce Intrvl B Beta
-.099126 -.609838
-.089645 -.630732
1.155752
LN HCCUT
LN COCUT
(Constant )
- .158962
- .140941
.981406
.028853
.024734
.084067
- .218799
- .192237
.807061
Variable T Sig T
LN_HCCUT -5.509 .0000
LN_COCUT -5.698 .0000
(Constant) 11.674 .0000
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_strt_
-> /METHOD=ENTER In hccut In cocut
**** MULTIPLE REGRESSION ***
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_STRT_ CO Strt ID
Block Number 1. Method: Enter LN HCCUT LN COCUT
M6IM001.WPD DRAFT 88 Mar 24, 1999
-------
DRAFT
Variable(s) Entered on Step Number
1.. LN_COCUT
2.. LN HCCUT
Multiple R .84999
R Square .72249
Adjusted R Square .69726
Standard Error .13266
Analysis of Variance
DF Sum of Squares Mean Square
Regression 2 1.00799 .50399
Residual 22 .38718 .01760
F = 28.63762 Signif F = .0000
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
LN_HCCUT -.159301 .032905 -.227541 -.091061 -.544428
LN_COCUT -.170728 .028208 -.229228 -.112229 -.680635
(Constant) 1.145947 .095873 .947118 1.344777
ln
Variable T Sig T
LN_HCCUT -4.841 .0001
LN_COCUT -6.053 .0000
(Constant) 11.953 .0000
M6IM001.WPD DRAFT 89 Mar 24, 1999
-------
DRAFT
APPENDIX G
Statistical Diagnostics for Running and Start High Emitter Levels
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & hc_2x = 2 & (grp88 = 3 | grp88 = 6 )).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 2 & (grp88 = 3 | grp88 = 6 )
-> ' (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> EXAMINE
-> VARIABLES=hc_cs hc_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
HC_CS
By FILTER_$ 1
Valid cases:
Selected
118.0 Missing cases:
Percent missing:
Mean 5.3127 Std Err .9562 Min
Median 3.8660 Variance 107.8864 Max
5% Trim 4.3032 Std Dev 10.3868 Range
95% CI for Mean (3.4190, 7.2064) IQR
-23.3000 Skewness 6.7474
100.5300 S E Skew .2227
123.8300 Kurtosis 61.6924
2.8755 S E Kurt .4419
HC_LA4HO HC_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
118.0 Missing cases:
Percent missing:
Mean 2.3725 Std Err .4448 Min
Median 1.0085 Variance 23.3486 Max
5% Trim 1.4788 Std Dev 4.8320 Range
95% CI for Mean (1.4916, 3.2535) IQR
.2690 Skewness 5.1217
34.8100 S E Skew .2227
34.5410 Kurtosis 28.7006
1.3385 S E Kurt .4419
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 2 & (grp88 = 3 | grp88 = 6 )).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 2 & (grp88 = 3 | grp88 = 6 )
-> ' (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXAMINE
-> VARIABLES=co_cs co_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
By
CO_CS
FILTER $
Selected
M6IM001.WPD DRAFT
90
Mar 24, 1999
-------
DRAFT
Valid cases:
97.0
Missing cases:
Percent missing:
Mean 65.3116 Std Err 9.4172 Min
Median 41.1230 Variance 8602.238 Max
5% Trim 63.1510 Std Dev 92.7483 Range
95% CI for Mean (46.6187, 84.0045) IQR
-181.100 Skewness .7955
441.8000 S E Skew .2450
622.9000 Kurtosis 2.4172
95.2160 S E Kurt .4853
CO_LA4HO CO_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
97.0
Missing cases:
Percent missing:
Mean 37.9327 Std Err 5.2679 Min
Median 14.1360 Variance 2691.801 Max
5% Trim 30.7305 Std Dev 51.8826 Range
95% CI for Mean (27.4761, 48.3894) IQR
.2920 Skewness 2.4569
288.6300 S E Skew .2450
288.3380 Kurtosis 6.7550
33.8540 S E Kurt .4853
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & no_2x = 2 & (grp88 = 3 | grp88 = 6 )).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 2 & (grp88 = 3 | grp88 = 6 )
-> ' (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> EXAMINE
-> VARIABLES=no_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
NO_LA4HO NO_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
44.0 Missing cases:
Percent missing:
.0
Mean 2.9513 Std Err .1349 Min
Median 2.5785 Variance .8006 Max
5% Trim 2.8761 Std Dev .8948 Range
95% CI for Mean (2.6793, 3.2233) IQR
1.9530 Skewness 1.2149
5.6660 S E Skew .3575
3.7130 Kurtosis .9399
1.2920 S E Kurt .7017
USE ALL.
COMPUTE filter_$=(vehicle = 1 & hc_2x = 2 &
7 ) ) .
VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 2
' grp88 = 7 ) (FILTER)'.
VALUE LABELS filter_$ 0
FORMAT filter_$ (fl.O).
FILTER BY filter_$.
(grp88 = 4 | grp88 = 5 | grp88 =
(grp88 = 4 | grp88 = 5 |
'Not Selected' 1 'Selected'
-> EXECUTE .
-> EXAMINE
-> VARIABLES=hc_cs hc_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
M6IM001.WPD DRAFT
91
Mar 24, 1999
-------
/NOTOTAL.
DRAFT
HC_CS
By FILTER_$ 1
Valid cases:
Selected
212.0 Missing cases:
Percent missing:
.0
Mean 10.5195 Std Err 1.6407 Min
Median 5.8390 Variance 570.6977 Max
5% Trim 7.8954 Std Dev 23.8893 Range
95% CI for Mean (7.2852, 13.7538) IQR
-5.3850 Skewness 11.1465
326.0100 S E Skew .1671
331.3950 Kurtosis 145.4885
6.8020 S E Kurt .3326
HC_LA4HO HC_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
212.0 Missing cases:
Percent missing:
.0
Mean 1.8447 Std Err .3111 Min
Median .7975 Variance 20.5202 Max
5% Trim 1.2606 Std Dev 4.5299 Range
95% CI for Mean (1.2314, 2.4580) IQR
.1390 Skewness 10.4292
59.8590 S E Skew .1671
59.7200 Kurtosis 129.3009
1.3962 S E Kurt .3326
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 2 & (grp88 = 4 | grp88 = 5 | grp88 =
-> 7 )) .
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 2 & (grp88 = 4 | grp88 = 5 |'+
-> ' grp88 = 7 ) (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> EXAMINE
-> VARIABLES=co_cs co_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
CO_CS
By FILTER_$ 1
Valid cases:
Selected
233.0 Missing cases:
Percent missing:
.0
Mean 92.8206 Std Err 5.4515 Min
Median 78.5740 Variance 6924.600 Max
5% Trim 88.8831 Std Dev 83.2142 Range
95% CI for Mean (82.0797, 103.5614) IQR
-145.000 Skewness .8815
401.0900 S E Skew .1595
546.0900 Kurtosis 1.8693
88.6325 S E Kurt .3176
CO_LA4HO CO_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
233.0 Missing cases:
Percent missing:
.0
Mean 27.6531 Std Err 2.7249 Min
Median 11.4820 Variance 1729.998 Max
5% Trim 21.1470 Std Dev 41.5932 Range
95% CI for Mean (22.2845, 33.0217) IQR
.1330 Skewness 3.2284
298.0400 S E Skew .1595
297.9070 Kurtosis 12.9400
21.1570 S E Kurt .3176
M6IM001.WPD DRAFT
92
Mar 24, 1999
-------
DRAFT
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & no_2x = 2 & (grp88 = 4 | grp88 = 5 | grp88 =
-> 7 ) ) .
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 2 & (grp88 = 4 | grp88 = 5 |'+
-> ' grp88 = 7 ) (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> EXAMINE
-> VARIABLES=no_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
NO_LA4HO NO_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
60.0
Missing cases:
Percent missing:
Mean 2.8719 Std Err .0991 Min
Median 2.6320 Variance .5898 Max
5% Trim 2.8139 Std Dev .7680 Range
95% CI for Mean (2.6735, 3.0703) IQR
1.8730 Skewness 1.2166
5.8210 S E Skew .3087
3.9480 Kurtosis 2.2562
1.1900 S E Kurt .6085
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & hc_2x = 2 & (n_group = 1 )).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 2 & (n_group = 1 ) (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> EXAMINE
-> VARIABLES=hc_cs hc_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
HC_CS
By FILTER_$ 1
Valid cases:
Selected
58.0 Missing cases:
Percent missing:
Mean 4.8290 Std Err .7673 Min
Median 3.9220 Variance 34.1484 Max
5% Trim 4.6639 Std Dev 5.8437 Range
95% CI for Mean (3.2925, 6.3655) IQR
-23.3000 Skewness -.7800
24.2470 S E Skew .3137
47.5470 Kurtosis 11.2352
3.3150 S E Kurt .6181
By
HC_LA4HO
FILTER_$
HC_LA4HOT
Selected
M6IM001.WPD DRAFT
93
Mar 24, 1999
-------
DRAFT
Valid cases:
58.0
Missing cases:
Percent missing:
Mean 1.7400 Std Err .5316 Min
Median .8790 Variance 16.3917 Max
5% Trim 1.1879 Std Dev 4.0487 Range
95% CI for Mean (.6754, 2.8045) IQR
.1450 Skewness 6.9800
31.1790 S E Skew .3137
31.0340 Kurtosis 51.3330
1.0667 S E Kurt .6181
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & hc_2x = 2 & (n_group = 2 )).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 2 & (n_group = 2 ) (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> EXAMINE
-> VARIABLES=hc_cs hc_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
HC_CS
By FILTER_$ 1
Valid cases:
Selected
38.0 Missing cases:
Percent missing:
.0
Mean 3.2927 Std Err .4646 Min
Median 3.1575 Variance 8.2028 Max
5% Trim 3.2990 Std Dev 2.8640 Range
95% CI for Mean (2.3513, 4.2341) IQR
-3.2350 Skewness -.0333
10.4150 S E Skew .3828
13.6500 Kurtosis .7728
3.0717 S E Kurt .7497
HC_LA4HO HC_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
38.0
Missing cases:
Percent missing:
.0
Mean 3.3937 Std Err 1.0523 Min
Median 1.5370 Variance 42.0754 Max
5% Trim 2.2123 Std Dev 6.4866 Range
95% CI for Mean (1.2616, 5.5257) IQR
.5030 Skewness 3.8432
34.8100 S E Skew .3828
34.3070 Kurtosis 15.8588
1.5170 S E Kurt .7497
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 2 & (n_group = 1 )).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 2 & (n_group = 1 ) (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> EXAMINE
-> VARIABLES=co_cs co_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
M6IM001.WPD DRAFT
94
Mar 24, 1999
-------
DRAFT
co_cs
By FILTER_$ 1
Valid cases:
Selected
44.0 Missing cases:
Percent missing:
.0
Mean 38.0579 Std Err 11.1153 Min
Median 33.6220 Variance 5436.158 Max
5% Trim 36.5122 Std Dev 73.7303 Range
95% CI for Mean (15.6419, 60.4740) IQR
-134.000 Skewness .5754
286.3600 S E Skew .3575
420.3600 Kurtosis 2.4815
73.3610 S E Kurt .7017
CO_LA4HO CO_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
44.0 Missing cases:
Percent missing:
.0
Mean 36.1057 Std Err 7.1383 Min
Median 19.5880 Variance 2242.051 Max
5% Trim 29.1347 Std Dev 47.3503 Range
95% CI for Mean (21.7099, 50.5015) IQR
5.0870 Skewness 3.8473
288.6300 S E Skew .3575
283.5430 Kurtosis 18.8008
30.8875 S E Kurt .7017
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 2 & (n_group = 2 )).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 2 & (n_group = 2 ) (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> EXAMINE
-> VARIABLES=co_cs co_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
CO_CS
By FILTER_$ 1
Valid cases:
Selected
43.0 Missing cases:
Percent missing:
Mean 27.1649 Std Err 14.4273 Min
Median 35.0280 Variance 8950.311 Max
5% Trim 33.9697 Std Dev 94.6061 Range
95% CI for Mean (-1.9505, 56.2804) IQR
-280.000 Skewness -1.3909
218.1000 S E Skew .3614
498.1000 Kurtosis 4.0345
74.8400 S E Kurt .7090
CO_LA4HO CO_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
43.0 Missing cases:
Percent missing:
Mean 46.5270 Std Err 8.1257 Min
Median 21.1950 Variance 2839.142 Max
5% Trim 40.2982 Std Dev 53.2836 Range
95% CI for Mean (30.1287, 62.9252) IQR
3.9840 Skewness 1.7022
216.8700 S E Skew .3614
212.8860 Kurtosis 2.4073
55.1990 S E Kurt .7090
M6IM001.WPD DRAFT
95
Mar 24, 1999
-------
DRAFT
-> USE ALL.
-> COMPUTE filter_$=(vehicle = 1 & no_2x = 2 & (n_group = 1 )).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 2 & (n_group = 1 ) (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> EXAMINE
-> VARIABLES=no_la4ho BY filter_$
-> /PLOT NONE
-> /STATISTICS DESCRIPTIVES
-> /CINTERVAL 95
-> /MISSING LISTWISE
-> /NOTOTAL.
NO_LA4HO NO_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
11.0 Missing cases:
Percent missing:
.0
Mean 2.8455 Std Err .3223 Min
Median 2.3870 Variance 1.1423 Max
5% Trim 2.7867 Std Dev 1.0688 Range
95% CI for Mean (2.1274, 3.5635) IQR
1.7130 Skewness .9851
5.0350 S E Skew .6607
3.3220 Kurtosis .0612
1.6230 S E Kurt 1.2794
USE ALL.
COMPUTE filter_$=(vehicle = 1 & no_2x = 2 & (n_group = 2 )).
VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 2 & (n_group = 2 ) (FILTER)
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (fl.O).
FILTER BY filter_$.
EXECUTE .
EXAMINE
VARIABLES=no_la4ho BY filter_$
/PLOT NONE
/STATISTICS DESCRIPTIVES
/CINTERVAL 95
/MISSING LISTWISE
/NOTOTAL.
NO_LA4HO NO_LA4HOT
By FILTER_$ 1 Selected
Valid cases:
15.0 Missing cases:
Percent missing:
.0
Mean 2.8723 Std Err .2612 Min
Median 2.4130 Variance 1.0235 Max
5% Trim 2.7682 Std Dev 1.0117 Range
95% CI for Mean (2.3121, 3.4326) IQR
1.9530 Skewness 1.9401
5.6660 S E Skew .5801
3.7130 Kurtosis 3.5993
.9110 S E Kurt 1.1209
M6IM001.WPD DRAFT
96
Mar 24, 1999
-------
DRAFT
APPENDIX H
Statistical Diagnostics for Running and Start Normal Emitter Levels
-> GET
-> FILE='D:\MOBILE6\IM\IM_CRED\NEW_CRED\EF5_DAT.SAV .
-> COMPUTE filter_$=(vehicle = 1 & hc_2x = 1 & grp88 = 1).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 1 & grp88 = 1 (FILTER)'.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_LA4HO HC_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .44218
R Square .19552
Adjusted R Square .19501
Standard Error .07163
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 1.97161 1.97161
Residual 1581 8.11227 .00513
F = 384.24703 Signif F = .0000
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .001385 7.0661E-05 .001247 .001524 .442178
(Constant) .021397 .003347 .014831 .027963
in
Variable T Sig T
MILEAGE 19.602 .0000
(Constant) 6.392 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
M6IM001.WPD DRAFT 97 Mar 24, 1999
-------
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
DRAFT
Multiple R .18470
R Square .03411
Adjusted R Square .03350
Standard Error .92659
Analysis of Variance
DF
Regression 1
Residual 1581
F =
55.83968
Sum of Squares
47.94271
1357.41150
Signif F = .0000
Mean Square
47.94271
.85858
Variable
MILEAGE
(Constant )
B SE B
.006830 9.1404E-04
1.998720 .043300
< n
H
95% Confdnce
.005037
1.913788
Intrvl B
.008623
2 .083652
Beta
.184701
Variable
MILEAGE
(Constant)
T Sig T
7.473
46.159
.0000
.0000
End Block Number
All requested variables entered.
-> COMPUTE filter_$=( vehicle = 1 & hc_2x = 1 & grp88 =2).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN( .05) POUT(.IO)
- > /NOORIGIN
-> /DEPENDENT hc_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable. . HC_LA4HO HC_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
M6IM001.WPD DRAFT
98
Mar 24, 1999
-------
Multiple R .54551
R Square .29758
Adjusted R Square .29596
Standard Error .07463
DRAFT
Analysis of Variance
DF
Regression 1
Residual 434
F =
183.86531
Sum of Squares
1.02400
2.41708
Signif F = .0000
Mean Square
1.02400
.00557
H
Variable
MILEAGE
(Constant)
H i
B
.001701
.004198
SE B
1.2544E-04
.007088
95% Confdnce Intrvl B
.001454 .001947
-.009733 .018128
Beta
.545510
Variable
T Sig T
MILEAGE 13.560 .0000
(Constant) .592 .5540
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .10853
R Square .01178
Adjusted R Square .00950
Standard Error .70086
Analysis of Variance
Regression
Residual
DF Sum of Squares
1 2.54088
434 213.18120
Signif F = .0234
Mean Square
2.54088
.49120
Variable
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.002679
1.901893
.001178 3.63926E-04
.066564 1.771064
.004995
2 .032721
Beta
.108529
in
M6IM001.WPD DRAFT
99
Mar 24, 1999
-------
DRAFT
Variable
MILEAGE
(Constant)
T Sig T
2.274
28.572
.0234
.0000
End Block Number
All requested variables entered.
-> COMPUTE filter_$=( vehicle = 1 & hc_2x = 1 & grp88 =3).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 1 & grp88 = 1 (FILTER) '
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN( .05) POUT(.IO)
- > /NOORIGIN
-> /DEPENDENT hc_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable. . HC_LA4HO HC_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable (s) Entered on Step Number
1 . . MILEAGE
Multiple R .34643
R Square .12001
Adjusted R Square .11859
Standard Error .12378
Analysis of Variance
DF
Regression 1
Residual 621
F =
84.69051
Sum of Squares
1.29756
9.51443
Signif F = .0000
Mean Square
1.29756
.01532
Variable
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.001439 1.5635E-04
.094216 .009189
.001132
.076171
Beta
.346426
in
Variable
T Sig T
MILEAGE 9.203 .0000
(Constant) 10.254 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
M6IMOO 1 . WPD DRAFT
1 00
Mar 24, 1 999
-------
DRAFT
/MISSING LISTWISE
/STATISTICS COEFF OUTS CI R ANOVA
/CRITERIA=PIN(.05) POUT(.IO)
/NOORIGIN
/DEPENDENT hc_cs
/METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .03793
R Square .00144
Adjusted R Square -.00017
Standard Error 1.16208
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 1.20815 1.20815
Residual 621 838.61274 1.35042
F = .89465 Signif F = .3446
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .001388 .001468 -.001494 .004271 .037929
(Constant) 2.358932 .086266 2.189523 2.528341
Variable T Sig T
MILEAGE .946 .3446
(Constant) 27.345 .0000
End Block Number 1 All requested variables entered.
- >
-> COMPUTE filter_$=(vehicle = 1 & hc_2x = 1 & grp88 = 4).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 1 & grp88 = 1 (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
M6IMOO1.WPD DRAFT 101 Mar 24, 1999
-------
DRAFT
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_LA4HO HC_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .25073
R Square .06286
Adjusted R Square .05245
Standard Error .09393
Analysis of Variance
Regression
Residual
DF Sum of Squares
1 .05327
90 .79414
Signif F = .0159
Mean Square
.05327
.00882
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
8.12421E-04 3.3064E-04 1.55544E-04 .001469
.077383 .020471 .036713 .118053
Beta
.250729
Variable T Sig T
MILEAGE 2.457 .0159
(Constant) 3.780 .0003
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .48064
R Square .23102
Adjusted R Square .22247
Standard Error .99649
Analysis of Variance
Regression
Residual
DF Sum of Squares
1 26.84774
90 89.36862
Mean Square
26.84774
.99298
F = 27.03742 Signif F = .0000
M6IM001.WPD DRAFT
102
Mar 24, 1999
-------
DRAFT
Variable
MILEAGE
(Constant )
B
.018238
1.493421
SE B
.003508
.217166
95% Confdnce
.011270
1.061982
Intrvl B
.025207
1.924860
Beta
.480640
MILEAGE
(Constant)
T Sig T
5.200 .0000
6.877 .0000
End Block Number
All requested variables entered.
-> COMPUTE filter_$=(vehicle = 1 & hc_2x = 1 & grp88 = 5).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_la4ho
-> /METHOD=ENTER mileage
MULTIPLE
REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_LA4HO HC_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .20394
R Square .04159
Adjusted R Square .03746
Standard Error .12495
Analysis of Variance
Regression
Residual
DF Sum of Squares
1 .15719
232 3.62209
Signif F = .0017
Mean Square
.15719
.01561
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.001214 3.8251E-04 4.60068E-04 .001967
.126577 .014947 .097128 .156026
Beta
203940
M6IM001.WPD DRAFT
103
Mar 24, 1999
-------
DRAFT
Variable T Sig T
MILEAGE 3.173 .0017
(Constant) 8.469 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .16213
R Square .02628
Adjusted R Square .02209
Standard Error 1.22801
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 9.44422 9.44422
Residual 232 349.86033 1.50802
F = 6.26267 Signif F = .0130
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .009408 .003759 .002001 .016815 .162126
(Constant) 1.589214 .146898 1.299790 1.878638
in
Variable T Sig T
MILEAGE 2.503 .0130
(Constant) 10.818 .0000
End Block Number 1 All requested variables entered.
- >
-> COMPUTE filter_$=(vehicle = 1 & hc_2x = 1 & grp88 = 6).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 1 & grp88 = 1 (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
M6IMOO1.WPD DRAFT 104 Mar 24, 1999
-------
DRAFT
/MISSING LISTWISE
/STATISTICS COEFF OUTS CI R ANOVA
/CRITERIA=PIN(.05) POUT(.IO)
/NOORIGIN
/DEPENDENT hc_la4ho
/METHOD=ENTER mileage
**** MULTIPLE REGRESSION **
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_LA4HO HC_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .50527
R Square .25529
Adjusted R Square .24806
Standard Error .11052
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 .43130 .43130
Residual 103 1.25812 .01221
F = 35.30979 Signif F = .0000
---------------------- Variables in the Equation -----------------------
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .002250 3.7865E-04 .001499 .003001 .505267
(Constant) .097024 .018871 .059598 .134450
Variable T Sig T
MILEAGE 5.942 .0000
(Constant) 5.141 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN( .05) POUT(.IO)
- > /NOORIGIN
-> /DEPENDENT hc_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable. . HC_CS
Block Number 1. Method: Enter MILEAGE
Variable (s) Entered on Step Number
1 . . MILEAGE
Multiple R .21032
M6IMOO 1 . WPD DRAFT 1 05 Mar 24, 1 999
-------
DRAFT
R Square .04423
Adjusted R Square .03495
Standard Error 1.14075
Analysis of Variance
DF
Regression 1
Residual 103
F =
4.76698
Sum of Squares
6.20337
134.03607
Signif F = .0313
Mean Square
6.20337
1.30132
MILEAGE
(Constant)
- Variables in the Equation
B SE B 95% Confdnce Intrvl B
.008533
2.354343
.003908 7.81969E-04
.194779 1.968044
.016284
2.740641
Beta
210319
MILEAGE
(Constant)
T Sig T
2.183 .0313
12.087 .0000
End Block Number
All requested variables entered.
-> COMPUTE filter_$=(vehicle = 1 & hc_2x = 1 & grp88 = 7).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_la4ho
-> /METHOD=ENTER mileage
MULTIPLE
REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_LA4HO HC_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .28245
R Square .07978
Adjusted R Square .07868
Standard Error .11606
Analysis of Variance
DF
Regression 1
Residual 837
F =
72 .56441
Sum of Squares
.97738
11.27372
Signif F = .0000
Mean Square
.97738
.01347
M6IM001.WPD DRAFT
106
Mar 24, 1999
-------
DRAFT
Variable
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.001271 1.4921E-04 9.78196E-04 .001564
.153943 .006278 .141621 .166266
Beta
.282452
in
Variable
MILEAGE
(Constant)
T Sig T
8.518 .0000
24.522 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT hc_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. HC_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .24760
R Square .06131
Adjusted R Square .06019
Standard Error 1.43178
Analysis of Variance
DF
Regression 1
Residual 837
F =
54.66519
Sum of Squares
112.06309
1715.84167
Signif F = .0000
Mean Square
112.06309
2.04999
- Variables in the Equation
B SE B 95% Confdnce Intrvl B
MILEAGE
(Constant)
.013610
2.121260
.001841
.077449
.009997
1.969242
.017224
2.273278
Beta
247602
Variable T Sig T
MILEAGE 7.394 .0000
(Constant) 27.389 .0000
End Block Number 1 All requested variables entered.
M6IM001.WPD DRAFT
107
Mar 24, 1999
-------
DRAFT
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 1 & grp88 = 1).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_LA4HO CO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .43497
R Square .18920
Adjusted R Square .18869
Standard Error 1.21705
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 549.21332 549.21332
Residual 1589 2353.66001 1.48122
F = 370.78421 Signif F = .0000
H
Variable
MILEAGE
(Constant )
H i
B
.022927
.458769
SE B
.001191
.056622
95% Confdnce Intrvl B
.020592 .025262
.347707 .569832
Beta
.434967
MILEAGE 19.256 .0000
(Constant) 8.102 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
M6IMOO1.WPD DRAFT 108 Mar 24, 1999
-------
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
DRAFT
Multiple R .01494
R Square .00022
Adjusted R Square -.00041
Standard Error 12.05858
Analysis of Variance
DF Sum of Squares
Regression 1 51.59410
Residual 1589 231055.54059
F = .35482 Signif F = .5515
Mean Square
51.59410
145.40940
H
Variable B
MILEAGE .007027
(Constant) 18.972536
-i n
SE B
.011797
.561015
95% Confdnce Intrvl B
-.016112 .030166
17.872129 20.072942
Beta
.014941
Variable
MILEAGE
(Constant)
T Sig T
.596
33.818
.5515
.0000
End Block Number
All requested variables entered.
-> COMPUTE filter_$=( vehicle = 1 & co_3x = 1 & grp88 =2).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN( .05) POUT(.IO)
- > /NOORIGIN
-> /DEPENDENT co_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable. . CO_LA4HO CO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
M6IM001.WPD DRAFT
109
Mar 24, 1999
-------
DRAFT
Multiple R .57491
R Square .33052
Adjusted R Square .32897
Standard Error 1.39129
Analysis of Variance
DF
Regression 1
Residual 431
F =
212.78325
Sum of Squares
411.88438
834.28637
Signif F = .0000
Mean Square
411.88438
1.93570
H
Variable
MILEAGE
(Constant)
B
.033909
- .028277
i n
SE B
.002325
.131686
95% Confdnce Intrvl B
.029340 .038478
-.287103 .230549
Beta
.574909
Variable
T Sig T
MILEAGE 14.587 .0000
(Constant) -.215 .8301
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .02245
R Square .00050
Adjusted R Square -.00182
Standard Error 8.95890
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 17.43774 17.43774
Residual 431 34592.88334 80.26191
F = .21726 Signif F = .6414
Variable B SE B 95% Confdnce Intrvl B
MILEAGE -.006977 .014969 -.036398 .022444
(Constant) 19.232859 .847958 17.566211 20.899506
-i n
Beta
- .022446
M6IM001.WPD DRAFT
110
Mar 24, 1999
-------
DRAFT
Variable T Sig T
MILEAGE -.466 .6414
(Constant) 22.681 .0000
End Block Number 1 All requested variables entered.
- >
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 1 & grp88 =3).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 1 & grp88 = 1 (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_LA4HO CO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .34381
R Square .11821
Adjusted R Square .11683
Standard Error 1.80541
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 279.20868 279.20868
Residual 639 2082.82364 3.25950
F = 85.65984 Signif F = .0000
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .019588 .002116 .015432 .023744 .343812
(Constant) 1.444769 .130120 1.189254 1.700284
in
Variable T Sig T
MILEAGE 9.255 .0000
(Constant) 11.103 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
M6IMOO1.WPD DRAFT 111 Mar 24, 1999
-------
DRAFT
/MISSING LISTWISE
/STATISTICS COEFF OUTS CI R ANOVA
/CRITERIA=PIN(.05) POUT(.IO)
/NOORIGIN
/DEPENDENT co_cs
/METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .01070
R Square .00011
Adjusted R Square -.00145
Standard Error 13.54016
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 13.41337 13.41337
Residual 639 117151.73251 183.33604
F = .07316 Signif F = .7869
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE -.004293 .015873 -.035462 .026875 -.010700
(Constant) 19.949338 .975872 18.033034 21.865642
Variable T Sig T
MILEAGE -.270 .7869
(Constant) 20.443 .0000
End Block Number 1 All requested variables entered.
- >
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 1 & grp88 = 4).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 1 & grp88 = 1 (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
M6IMOO1.WPD DRAFT 112 Mar 24, 1999
-------
DRAFT
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_LA4HO CO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .38730
R Square .15000
Adjusted R Square .14076
Standard Error 1.02382
Analysis of Variance
Regression
Residual
DF Sum of Squares
1 17.01790
92 96.43454
Signif F = .0001
Mean Square
17.01790
1.04820
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.013709 .003402 .006952 .020467
.566553 .216869 .135832 .997274
Beta
.387299
Variable T Sig T
MILEAGE 4.029 .0001
(Constant) 2.612 .0105
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .16121
R Square .02599
Adjusted R Square .01540
Standard Error 21.02393
Analysis of Variance
Regression
Residual
DF
1
92
Sum of Squares
1085.07366
40664.53507
Mean Square
1085.07366
442.00582
Signif F = .1206
M6IM001.WPD DRAFT
113
Mar 24, 1999
-------
DRAFT
Variable
MILEAGE
(Constant )
B
.109470
24.697606
SE B
.069868
4.453376
95% Confdnce
- .029294
15.852816
Intrvl B
.248234
33 .542395
Beta
.161214
MILEAGE
(Constant)
T Sig T
1.567 .1206
5.546 .0000
End Block Number
All requested variables entered.
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 1 & grp88 = 5).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_la4ho
-> /METHOD=ENTER mileage
MULTIPLE
REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_LA4HO CO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .24144
R Square .05829
Adjusted R Square .05423
Standard Error 1.46214
Analysis of Variance
Regression
Residual
DF Sum of Squares
1 30.70190
232 495.98217
Signif F = .0002
Mean Square
30.70190
2.13785
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.016908 .004462 .008118 .025699
.727606 .175115 .382587 1.072626
Beta
241439
M6IM001.WPD DRAFT
114
Mar 24, 1999
-------
DRAFT
Variable T Sig T
MILEAGE 3.790 .0002
(Constant) 4.155 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .10909
R Square .01190
Adjusted R Square .00764
Standard Error 20.73715
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 1201.49589 1201.49589
Residual 232 99766.81401 430.02937
F = 2.79399 Signif F = .0960
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .105775 .063281 -.018903 .230453 .109086
(Constant) 24.442451 2.483616 19.549126 29.335775
in
Variable T Sig T
MILEAGE 1.672 .0960
(Constant) 9.841 .0000
End Block Number 1 All requested variables entered.
- >
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 1 & grp88 = 6).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 1 & grp88 = 1 (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
M6IMOO1.WPD DRAFT 115 Mar 24, 1999
-------
DRAFT
/MISSING LISTWISE
/STATISTICS COEFF OUTS CI R ANOVA
/CRITERIA=PIN(.05) POUT(.IO)
/NOORIGIN
/DEPENDENT co_la4ho
/METHOD=ENTER mileage
**** MULTIPLE REGRESSION **
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_LA4HO CO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .32292
R Square .10428
Adjusted R Square .09583
Standard Error 1.75396
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 37.96399 37.96399
Residual 106 326.09694 3.07639
F = 12.34045 Signif F = .0007
---------------------- Variables in the Equation -----------------------
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .021502 .006121 .009367 .033637 .322923
(Constant) 1.576249 .300873 .979739 2.172759
Variable T Sig T
MILEAGE 3.513 .0007
(Constant) 5.239 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN( .05) POUT(.IO)
- > /NOORIGIN
-> /DEPENDENT co_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable. . CO_CS
Block Number 1. Method: Enter MILEAGE
Variable (s) Entered on Step Number
1 . . MILEAGE
Multiple R .27648
M6IMOO 1 . WPD DRAFT 1 1 6 Mar 24, 1 999
-------
DRAFT
R Square .07644
Adjusted R Square .06773
Standard Error 25.80196
Analysis of Variance
DF
Regression 1
Residual 106
F =
S.77340
Sum of Squares
5840.81033
70568.54286
Signif F = .0038
Mean Square
5840.81033
665.74097
Variable
MILEAGE
(Constant )
B
.266706
20.038190
SE B
.090043
4.426039
95% Confdnce
.088187
11.263137
Intrvl B
.445224
28.813243
Beta
.276480
MILEAGE
(Constant)
T Sig T
2.962 .0038
4.527 .0000
End Block Number 1 All requested variables entered.
- >
-> COMPUTE filter_$=(vehicle = 1 & co_3x = 1 & grp88 = 7).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_la4ho
-> /METHOD=ENTER mileage
MULTIPLE
REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_LA4HO CO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .20587
R Square .04238
Adjusted R Square .04121
Standard Error 1.75821
Analysis of Variance
DF
Regression 1
Residual 814
F =
36.02576
Sum of Squares
111.36702
2516.33162
Signif F = .0000
Mean Square
111.36702
3.09132
M6IM001.WPD DRAFT
117
Mar 24, 1999
-------
DRAFT
Variable
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.013887
1.393200
.002314
.096173
.009346
1.204424
.018429
1.581977
Beta
.205869
in
Variable
MILEAGE
(Constant)
T Sig T
6.002 .0000
14.486 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT co_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. CO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .23702
R Square .05618
Adjusted R Square .05502
Standard Error 24.75160
Analysis of Variance
DF
Regression 1
Residual 814
F =
48.45262
Sum of Squares
29684.08673
498690.22165
Signif F = .0000
Mean Square
29684.08673
612.64155
- Variables in the Equation
B SE B 95% Confdnce Intrvl B
MILEAGE
(Constant)
.226728
28.636627
.032572
1.353894
.162792
25.979092
.290663
31.294161
Beta
.237023
Variable T Sig T
MILEAGE 6.961 .0000
(Constant) 21.151 .0000
End Block Number 1 All requested variables entered.
M6IM001.WPD DRAFT
118
Mar 24, 1999
-------
DRAFT
-> COMPUTE filter_$=(vehicle = 1 & no_2x = 1 & grp88 = 1).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 1 & grp88 = 1 (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT no_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_LA4HO NO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .39011
R Square .15219
Adjusted R Square .15166
Standard Error .22872
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 15.10851 15.10851
Residual 1609 84.16805 .05231
F = 288.82214 Signif F = .0000
---------------------- Variables in the Equation -----------------------
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .003756 2.2100E-04 .003322 .004189 .390110
(Constant) .200589 .010576 .179844 .221334
Variable T Sig T
MILEAGE 16.995 .0000
(Constant) 18.966 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN( .05) POUT(.IO)
- > /NOORIGIN
-> /DEPENDENT no_cs
-> /METHOD=ENTER mileage
MULTIPLE REGRESSION
M6IMOO1.WPD DRAFT 119 Mar 24, 1999
-------
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
DRAFT
Multiple R .05934
R Square .00352
Adjusted R Square .00290
Standard Error .95537
Analysis of Variance
DF
Regression 1
Residual 1609
F =
5.68559
Sum of Squares
5.18937
1468.57163
Signif F = .0172
Mean Square
5.18937
.91272
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.002201 9.2312E-04 3.90488E-04 .004012
1.443620 .044179 1.356966 1.530274
Beta
.059340
MILEAGE
(Constant)
T Sig T
2.384 .0172
32.677 .0000
End Block Number 1 All requested variables entered.
- >
-> COMPUTE filter_$=( vehicle = 1 & no_2x = 1 & grp88 =2).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN( .05) POUT(.IO)
- > /NOORIGIN
-> /DEPENDENT no_la4ho
-> /METHOD=ENTER mileage
MULTIPLE
REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable. . NO_LA4HO NO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
M6IM001.WPD DRAFT
120
Mar 24, 1999
-------
Multiple R .44967
R Square . 20220
Adjusted R Square .20038
Standard Error .21488
DRAFT
Analysis of Variance
Regression
Residual
111.26452
DF
1
439
Sum of Squares
5.13724
20.26926
Signif F = .0000
Mean Square
5.13724
.04617
H
Variable
MILEAGE
(Constant )
H i
B
.003806
.225302
SE B
3 .6084E-04
.020306
95% Confdnce Intrvl B
.003097 .004515
.185393 .265212
Beta
.449669
Variable
T Sig T
MILEAGE 10.548 .0000
(Constant) 11.095 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT no_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .10744
R Square .01154
Adjusted R Square .00929
Standard Error 1.17515
Analysis of Variance
DF
Regression 1
Residual 439
F =
5.12705
Sum of Squares
7.08035
606.25000
Signif F = .0240
Mean Square
7.08035
1.38098
MILEAGE
(Constant)
- Variables in the Equation
B SE B 95% Confdnce Intrvl B
-.004468
2.300454
.001973
.111055
-.008347 -5.89886E-04
2.082188 2.518720
Beta
- .107444
M6IM001.WPD DRAFT
121
Mar 24, 1999
-------
DRAFT
Variable
MILEAGE
(Constant)
T Sig T
-2.264
20.715
.0240
.0000
End Block Number
All requested variables entered.
-> COMPUTE filter_$=( vehicle = 1 & no_2x = 1 & grp88 =3).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN( .05) POUT(.IO)
- > /NOORIGIN
-> /DEPENDENT no_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable. . NO_LA4HO NO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .18619
R Square .03467
Adjusted R Square .03327
Standard Error .33374
Analysis of Variance
DF
Regression 1
Residual 692
F =
24.85020
Sum of Squares
2.76791
77.07747
Signif F = .0000
Mean Square
2.76791
.11138
H
Variable
MILEAGE
(Constant)
H i
B
.001883
.479830
SE B
3 .7774E-04
.023534
95% Confdnce Intrvl B
.001141 .002625
.433623 .526037
Beta
.186188
Variable
T Sig T
MILEAGE 4.985 .0000
(Constant) 20.389 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
M6IMOO 1 . WPD DRAFT
1 22
Mar 24, 1 999
-------
DRAFT
/MISSING LISTWISE
/STATISTICS COEFF OUTS CI R ANOVA
/CRITERIA=PIN(.05) POUT(.IO)
/NOORIGIN
/DEPENDENT no_cs
/METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .04406
R Square .00194
Adjusted R Square .00050
Standard Error 1.07313
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 1.54972 1.54972
Residual 692 796.91339 1.15161
F = 1.34570 Signif F = .2464
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .001409 .001215 -9.75750E-04 .003794 .044055
(Constant) 1.406422 .075673 1.257846 1.554999
Variable T Sig T
MILEAGE 1.160 .2464
(Constant) 18.586 .0000
End Block Number 1 All requested variables entered.
- >
-> COMPUTE filter_$=(vehicle = 1 & no_2x = 1 & grp88 = 4).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 1 & grp88 = 1 (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT no_la4ho
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION ****
M6IMOO1.WPD DRAFT 123 Mar 24, 1999
-------
DRAFT
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_LA4HO NO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .16411
R Square .02693
Adjusted R Square .01647
Standard Error .32441
Analysis of Variance
Regression
Residual
DF Sum of Squares
1 .27088
93 9.78756
Signif F = .1120
Mean Square
.27088
.10524
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.001702 .001061 -4.04579E-04 .003808
.495967 .067974 .360984 .630950
Beta
.164107
Variable T Sig T
MILEAGE 1.604 .1120
(Constant) 7.296 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT no_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .06506
R Square .00423
Adjusted R Square -.00647
Standard Error 1.05703
Analysis of Variance
Regression
Residual
DF
1
93
Sum of Squares
.44169
103.91026
Mean Square
.44169
1.11731
Signif F = .5311
M6IM001.WPD DRAFT
124
Mar 24, 1999
-------
DRAFT
Variable
MILEAGE
(Constant )
B
- .002173
1.404902
SE B
.003456
.221481
95% Confdnce
- .009035
.965084
Intrvl B
.004690
1.844719
Beta
- .065059
MILEAGE
(Constant)
T Sig T
-.629 .5311
6.343 .0000
End Block Number
All requested variables entered.
-> COMPUTE filter_$=(vehicle = 1 & no_2x = 1 & grp88 = 5).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT no_la4ho
-> /METHOD=ENTER mileage
MULTIPLE
REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_LA4HO NO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .15950
R Square .02544
Adjusted R Square .02148
Standard Error .38404
Analysis of Variance
Regression
Residual
DF Sum of Squares
1 .94709
246 36.28266
Signif F = .0119
Mean Square
.94709
.14749
MILEAGE
(Constant)
Variables in the Equation
B SE B 95% Confdnce Intrvl B
.002725 .001075 6.06968E-04 .004843
.555546 .044174 .468539 .642552
Beta
.159497
M6IM001.WPD DRAFT
125
Mar 24, 1999
-------
DRAFT
Variable T Sig T
MILEAGE 2.534 .0119
(Constant) 12.576 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT no_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .10136
R Square .01027
Adjusted R Square .00625
Standard Error 1.17091
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 3.50127 3.50127
Residual 246 337.27551 1.37104
F = 2.55374 Signif F = .1113
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .005240 .003279 -.001218 .011698 .101363
(Constant) .747776 .134681 .482502 1.013050
in
Variable T Sig T
MILEAGE 1.598 .1113
(Constant) 5.552 .0000
End Block Number 1 All requested variables entered.
- >
-> COMPUTE filter_$=(vehicle = 1 & no_2x = 1 & grp88 = 6).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 1 & grp88 = 1 (FILTER)'.
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
M6IMOO1.WPD DRAFT 126 Mar 24, 1999
-------
DRAFT
/MISSING LISTWISE
/STATISTICS COEFF OUTS CI R ANOVA
/CRITERIA=PIN(.05) POUT(.IO)
/NOORIGIN
/DEPENDENT no_la4ho
/METHOD=ENTER mileage
**** MULTIPLE REGRESSION **
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_LA4HO NO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .41146
R Square .16930
Adjusted R Square .16146
Standard Error .38335
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 3.17478 3.17478
Residual 106 15.57778 .14696
F = 21.60302 Signif F = .0000
---------------------- Variables in the Equation -----------------------
Variable B SE B 95% Confdnce Intrvl B Beta
MILEAGE .006326 .001361 .003628 .009024 .411459
(Constant) .459727 .066906 .327079 .592376
Variable T Sig T
MILEAGE 4.648 .0000
(Constant) 6.871 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN( .05) POUT(.IO)
- > /NOORIGIN
-> /DEPENDENT no_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable. . NO_CS
Block Number 1. Method: Enter MILEAGE
Variable (s) Entered on Step Number
1 . . MILEAGE
Multiple R .01562
M6IMOO 1 . WPD DRAFT 1 27 Mar 24, 1 999
-------
DRAFT
R Square .00024
Adjusted R Square -.00919
Standard Error 1.03573
Analysis of Variance
DF
Regression 1
Residual 106
F =
.02587
Sum of Squares
.02775
113.71076
Signif F = .8725
Mean Square
.02775
1.07274
Variables in the Equation
Variable B SE B 95% Confdnce Intrvl B
MILEAGE 5.91396E-04 .003677 -.006699 .007882
(Constant) 1.530162 .180765 1.171777 1.888547
Beta
.015619
MILEAGE
(Constant)
T Sig T
.161 .8725
8.465 .0000
End Block Number
All requested variables entered.
-> COMPUTE filter_$=(vehicle = 1 & no_2x = 1 & grp88 = 7).
-> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x = 1 & grp88 = 1 (FILTER)
-> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
-> FORMAT filter_$ (fl.O).
-> FILTER BY filter_$.
-> EXECUTE .
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT no_la4ho
-> /METHOD=ENTER mileage
MULTIPLE
REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_LA4HO NO_LA4HOT
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .17186
R Square .02954
Adjusted R Square .02854
Standard Error .37012
Analysis of Variance
DF
Regression 1
Residual 972
F =
29.58196
Sum of Squares
4.05244
133.15445
Signif F = .0000
Mean Square
4.05244
.13699
M6IM001.WPD DRAFT
128
Mar 24, 1999
-------
DRAFT
Variable
MILEAGE
(Constant)
Variables in the Equation
95% Confdnce Intrvl B
.002328 4.2806E-04
.583430 .019264
.001488
.545626
.003168
.621234
Beta
.171858
in
Variable
MILEAGE
(Constant)
T Sig T
5.439 .0000
30.286 .0000
End Block Number 1 All requested variables entered.
-> REGRESSION
-> /MISSING LISTWISE
-> /STATISTICS COEFF OUTS CI R ANOVA
-> /CRITERIA=PIN(.05) POUT(.IO)
-> /NOORIGIN
-> /DEPENDENT no_cs
-> /METHOD=ENTER mileage
**** MULTIPLE REGRESSION
Listwise Deletion of Missing Data
Equation Number 1 Dependent Variable.. NO_CS
Block Number 1. Method: Enter MILEAGE
Variable(s) Entered on Step Number
1.. MILEAGE
Multiple R .17750
R Square .03151
Adjusted R Square .03051
Standard Error 1.36097
Analysis of Variance
DF
Regression 1
Residual 972
F =
31.61932
Sum of Squares
58.56677
1800.38342
Signif F = .0000
Mean Square
58.56677
1.85225
MILEAGE
(Constant)
- Variables in the Equation
B SE B 95% Confdnce Intrvl B
-.008851
1.601358
.001574
.070836
-.011940
1.462349
-.005762
1.740366
Beta
- .177497
Variable T Sig T
MILEAGE -5.623 .0000
(Constant) 22.607 .0000
M6IM001.WPD DRAFT
129
Mar 24, 1999
-------
DRAFT
Figure2-Biennial II with BSD
2 i 4
P 9 10 ii 12. i'i 14-
M6IM001.WPD DRAFT
130
Mar 24, 1999
------- |