ANALYSIS OF SOLID WASTE COMPOSITION


      Statistical Technique to Determine

                 Sample Size

                   SW-19ts


     Dennis E. Carruth and Albert J. Klee
 U.S.. DEPARTMENT OF HEALTH,  EDUCATION,  AND WELFARE
                Public Health Service
Consumer Protection and Environmental Health Service
        Environmental  Control Administration
          Bureau of Solid Waste Management
                         1969

-------
Single copies of this publication will be distributed as supplies
permit.  Request from the Bureau of Solid Waste Management, Envi-
ronmental Control Administration, 5555 Ridge Avenue, Cincinnati,
Ohio  J»5213.

-------
                                ABSTRACT






     This work analyzes data obtained after separating and weighing




solid waste and presents a statistical technique to determine the




minimum weight and number of samples needed to realistically and




reliably estimate the characteristics of a given quantity of solid




waste.  Samples of solid waste of varying weights were taken and




separated into nine components; each component was weighed, and the




data were then statistically evaluated.   It was determined that there




is no significant variance among sample weight groups.  The study




recommends that 12, relatively small (200-lb)  samples will validly




reflect the composition of a given supply of solid waste.  (Author)

-------

-------
                                CONTENTS

                                                                   PAGE

INTRODUCTION 	     1

     Solid Waste Composition Criteria  	     2

     Sample Criteria 	     4


DISCUSSION 	     7

     Incineration Studies  	     7

     Solid Waste Composition Classification  	     8

     Method of Taking Samples  	     9

     Effect of Sample Weight on Precision  	   JQ

     Statistical Technique to Determine Sample Size  	   17

        Example	   18


SUMMARY AND CONCLUSIONS  	   23


REFERENCES	   25
TABLES
     1   Weight, In Pounds, of Components Per Sample
        and Per Weight Group	    11

     2  Percentage (by weight)  of Components Per Sample
        and Per Weight Group	    13

     3  Critical Statistics Obtained from Table 2  	    15

     k  Coefficients of Variation  	    16

     5  Arcsin Transformation  	    ig

-------

-------
                  ANALYSIS OF SOLID WASTE COMPOSITION

            Statistical Technique  to Determine Sample Size


                 Dennis E. Carruth and Albert J. Klee*


                               INTRODUCTION
     Whenever attempts are made to advance the frontiers of knowledge

and to acquire additional understanding of a phenomenon, operation,

or system, hypotheses must be formulated and tested.  Certainly, the

data gathered in the course of such studies must contribute to estab-

lished objectives.  Such a process of basic and applied research

(reporting ->• description -> explanation)1 is universal to all creative

disciplines and research endeavors.

     The Bureau of Solid Waste Management is not unique in this

respect nor is it a stranger to scientific research  in attempting to

broaden the existing horizons of knowledge in solid waste management.

One objective is for Federal, State, county, and municipal governments,

industry, and other interested and concerned institutions and indivi-

duals to obtain increased knowledge and better understanding of solid

waste pract ices.
    *Mr. Carruth is Staff Engineer, Division of Research and Development,
and Mr. Klee is Chief, Operational Analysis Branch, Division of Technical
Operations of the Bureau of Solid Waste Management, Environmental Control
Administration, Consumer Protection and Environmental Health Service,
Public Health Service, U.S. Department of Health, Education, and Wel-
fare.

-------
     Solid waste is defined by the Solid Waste Disposal Act of 1965

as:

     .  .  .garbage, refuse, and other discarded solid materials,
     including solid waste materials resulting from indus-
     trial, commercial, and agricultural operations, and from
     community activities, but does not include solids or
     dissolved material in domestic sewage or other signif-
     icant pollutants in water resources, such as silt, dis-
     solved or suspended solids in industrial waste water
     effluents, dissolved materials in  irrigation return flows
     or other common water pollutants.2

     To broaden the base of understanding in solid waste management,

the  Bureau's Division of Technical Operations is charged with the re-

sponsibility to provide technical  assistance, to develop basic technical

information on solid wastes and solid waste management practices, to

manage the State planning grant program, and to apply management

science to problems of solid waste handling and disposal.  The commonly

recognized methods of disposal, which can prove satisfactory, include

sanitary landfilling, incinerating, and composting, as well as sal-

vage and reuse operations.


                    Solid Waste Composition Criteria


     Through legislative action and increased public awareness, con-

siderable  interest and activity have recently developed in the solid

waste field.  Solid waste research, however, is primarily speculative

in nature and  is currently more concerned with gathering data and

establishing hypotheses than with formulating broad generalizations

or conclusions.

-------
     An  important phase  in data collection and survey design  is




quantifying and qualifying solid waste materials generated from a




multitude of sources,  including residential, commercial,  industrial,




and  institutional origins.  Yet, our knowledge of the physical and




chemical composition of  solid waste has remained undeveloped.  Solid




waste  is not only complex in  its composition, but also extremely




variable because of such factors as geographical location, season




of the year, social and  economic conditions, and methods and  fre-




quency of collection.




     For example, to design and operate a municipal incinerator and




to evaluate its efficiency in processing solid waste, the Btu (heat)




content of the fuel, i.e., solid waste, must be determined.   Sanitary




landfill, as a disposal  method, requires a knowledge of solid waste




composition to determine compaction properties, problems of leachates,




gaseous emissions, and so forth.  Composting requires the separation




of glass, metals, rocks, and other nondegradable items from the




otherwise compostible material.




     To  learn more about the nature of solid waste, a reliable and




standard sampling procedure must be developed so that meaningful




samples of solid waste can be obtained for analysis and study.  Studies




at Purdue University determined that analyzing a quarter-truckload




of solid waste produced  satisfactory data.3  From this, the concept




of quartering a given supply of solid waste before appropriate anal-




ysis and separation was developed.   Further, a statistical analysis

-------
gave the approximate number of sample households, in a homogeneous

area of a community, needed to estimate such quantities as pounds

per capita per day, pounds per household per day, cubic feet per

capita per day, and bulk density.  When solid wastes generated from

a multitude of sources are analyzed, however, these concepts do

not seem to apply.


                             Sample Cr i teria


     The preceding discussion indicates the need for quantifying

and classifying solid waste materials.  It is, of course, practically

impossible and physically undesirable to separate, measure, weigh,

and analyze all of the solid waste deposited at a particular facility.

Therefore, should all the solid waste in a storage pit of an Incin-

erator be separated and analyzed, or what portion thereof?  In de-

termining the physical composition of a truckload of solid waste,

should the entire truckload be sampled, or rather a smaller fraction?

What weight sample yields reasonable results and how many such sam-

ples are needed?  Obtaining answers to these and similar questions

is troublesome, yet important.

     Samples must be taken in sufficient weight and number to realis-

tically and reliably estimate the characteristics of a given quantity

of sol id waste.

     In the planning of a sample survey, a stage is always
     reached at which a decision must be made about the size
     of the sample.  The decision is important.  Too large a

-------
     sample implies a waste of resources, and too small a
     sample diminishes the utility of the results.  The de-
     cision cannot always be made satisfactorily, for often
     we do not possess enough information to be sure that
     our choice of sample size is the best one.1*

     Previous studies have not fully treated the problem of obtaining

adequate samples of solid waste for analysis, nor have practices been

developed to the extent of producing statistically comparable data.

The variable composition of solid waste from a given source should

be definable, however, and the following discussion, though specific

in scope, should offer a general  technique by which to analyze solid

waste composition.

-------

-------
                               DISCUSSION






                           Incineration Studies






     These studies, which are projects of considerable depth and




scope, are undertaken by the Bureau's Division of Technical Operations




upon request of appropriate State, local, and private concerns.  Data




are provided for comparative and operational purposes and, hopefully,




for future recommendations.




     The study utilizes a systems approach whereby the total process




of incineration is observed, tested, and evaluated.  Collection trucks




and vehicles are weighed to provide true weights of the quantity of




solid waste handled by the  incinerator.  Samples of solid waste mate-




rial from the storage pit are taken, separated, and weighed to provide




data on Btu content and composition of the input.  Study of the in-




cineration process also yields information on mechanical function,




burning rate, burning temperature, pollution control  systems, feed




flow,  normal  and unique situations and conditions, and operating




and control procedures.  Samples of residue, gaseous and particulate




emissions, and liquid effluents are taken to quantify and qualify




the output of the systems and to ascertain the efficiency of the in-




cineration process.  Physical plant and disposal facilities, as well




as economic and administrative support facilities, are also studied




and evaluated.

-------
                 Solid Waste Composition Classification






     Determining the physical and quantitative composition of the




input—solid waste—is a necessary step in studying the acceptability




of incineration as a disposal method.  The composition of the solid




waste directly influences the Btu content of the waste, and deter-




mining the Btu content is necessary to evaluate the heat and com-




bustion properties of this feed material.




     After accounts of firsthand experiences and available literature




were analyzed and reviewed, solid wastes were classified into nine




categories.  Each category is composed of materials similar in compo-




sition and Btu content; each category permits relatively direct com-




parisons with existing data; and each category is composed of materials




that can be effectively separated, when the human element and repro-




ducible methodology are considered.



     The standard classification, based on nine mutually exclusive




components, is (1) food waste;  (2) garden waste; (3) paper products;




(k) plastic, rubber, and leather;  (5) textiles; (6) wood; (7) metals;




(8) glass and ceramics; (9) ash, rocks, and dirt (inerts).




     The classification is adaptable to analyze waste received at any




disposal facility, although there are several significant differences




among various methods of solid waste handling (e.g., whether or not




"bulky" wastes, such as discarded  refrigerators, are accepted).

-------
                        Method of Taking Samples






     For 4 consecutive days, the study surveyed and collected data




from an incineration facility that primarily handled residential and




commercial solid wastes.  A more thorough and complete study might




conceivably evaluate the composition of solid waste based upon




daily, weekly, monthly, seasonal, and yearly generation cycles.




     Statistical analyses (in particular, Analysis of Variance) have




shown that often there are significant differences in composition




among solid waste samples taken on the same day and on different




days during the week.  This additional source of variability is,




of course, unwanted  in any experiment designed to compare precisions




of samples of different weights.  Unfortunately, securing a sufficient




quantity of "homogeneous" material  to perform such studies is impossi-




ble.  Each truckload of material arriving at a disposal site represents




a different population of waste.  Indeed, much the same occurs with




different grapple loads of waste taken from an incinerator pit even




though attempts are usually made to mix the material.  Thus, to re-




duce such sources of variation, samples of different weights must be




taken at random times and days ("random block designs") or a system-




atic sampling plan ("partially random block designs") must be employed




to achieve the same results.  In this study, the latter course was




followed so that samples of a particular weight class were partially




randomized with regard to time and  day.  Those effects  inflate the




variances of the subsequent analyses, but they do not bias the varia-




tion of any weight class.

-------
     Past experience indicates that four men can reasonably separate




and weigh approximately 2,000 Ib of solid waste material in an aver-




age 8-hr day.  The decision whether to analyze one 2,000-lb sample




each day or whether analyzing smaller samples would be more reliable




was resolved by taking approximately equal numbers of samples in




three size ranges:  Weight Group I, 1,400 to 1,700 Ib; Weight Group II,




700 to 900 Ib; and Weight Group III, 200 to 300 Ib.  During the study,




a total of 11 such samples were taken — three in the largest size range




and four each in the smaller ranges (Table 1).




     The samples were taken with a front-end loader from the incoming




waste storage pile.  Particular care was taken to obtain samples




similar to the bulk of the waste being delivered.   To avoid biasing




the samples by actually handpicking the material to be classified,




however, each sample was removed in a single scoop from a point in




the pile that appeared to contain a relatively complete mixture of




the waste being delivered.  Each sample scoop was  subsequently placed




on a cleaned area of the dumping floor and immediately hand-separated




into the nine categories specified.  The weight of each component was




then recorded.






                   Effect of Sample Weight on Precision






     The study plan recognized the necessity of selecting samples of




solid waste of varying weights.  The amount of time and manpower avail-




able, of course,  limited the number of samples taken.  Of primary
                                   10

-------
 o
 ac
 o
UJ


cc
LU
a.

o
to

DC
UJ
a.
o
a.

o
o
00
a
o
o.
          a
          3
          O
          a
c
4)
C
O
a
g
0
ra
4-*
0


V
N
•_
I/I

V
1




"jj

o


V
N
in

1)
I





u
4-*


O
N

M
V
i












tfl
(D
O





O
o


F—
m
eg

O
r--
CM
o





I

LA
en

f*-


58
o
en



2


IA
»
*
8








4-*
£
u
LA
eg
LA


VO


CM
*~


m



vO


LA
m
LA
cn

CM

CO
m

LA
\o
CM



vO
S

O
S

*


O
VO

2







4)
4J
S
s
I
LA
VO


-3-
CM

cn



LA



VO


-
LA
en
o


CM

LA
IA



LA

—

O
CN

S


o
S

2







t)
4J
(A
J
c
•g
3
0
CO
vO
-3-
LA

CM

vO
CO
—


fV.



^

"
g
LA
LA
LA
0
CM
LA
LA

LA
CM

-3-


5
cn
LA
LA
O
*
CM"



IA
S

§







VI
4J
U
a

S.
10
a.
o
CO
CM


-3-


VO



CM



O


-
LA
en



cn

LA
VO




CM
CM

LA
g

^


IA
e

CM
L.
£

jf
«


rubber,
o

4J
M
LA
CM


^
CNJ

IA
*"


^^



en


CM
LA
en



-

LA
^.
CM




CM

O
CM

en


o
CM

cn








jj
^~
jjj
o
cn





vO



CM



en


CO
LA
O

—

CM
LA

LA
VO
CM


LA

CM

LA
vg

CN


LA
CM

CO










|
O
S
r-.



-------
importance, surely, is to determine the reliability and precision of




selecting a specified number of samples of a certain weight.




     To evaluate the original data, the weights of each component were




converted to percentages of the total  weight of the sample (Table 2).




By grouping the data according to sample weight ranges, comparisons




can be made among all  samples, among samples within groups, and among




groups.  Such analysis leads to the ultimate objective of determining




the optimum weight and number of samples to be taken.  By presenting




the data in percentage form, the percentage of a given component can be




compared among samples and groups and among all components within a




given sample or group.  It  is the variances of these percentages, how-




ever, that determine the sample weight and number of samples  needed to




adequately determine the composition of a given source of solid waste.




     The data expressed as percentages follow a multinomial distribution.




Before attempting statistical analysis, however, it is important to




note that the individual percentages cannot, in general,  be expected




to follow a normal distribution.  A normal distribution is considered




a good approximation to the distribution of a percentage  when the




percentage lies between 30 and 70 percent.  Only the paper component




meets this criterion (Table 2); the other entries fall below the minimum




acceptable 30 percent.




     The following is an excellent normalizing transformation, however.









                            Y = 2 arcsin /X
                                   12

-------










o_

o
f£
u
H-
u>
LU
•3-
DC
LU
a.
a
z
<
QJ
_J
Q-
5"
•S
<.
CO
Q£
LU
a.
CM
CO
LU H-
i z
CO LU
< Z
1- 0
a.
a:
o
c_>
u.
0
"p
t3
*-*
1 1 1

>-
03


LU
O
ff
H
Z
LU
O
flC
LU
a.





















"^

a


u
19
4J
.C
Ol
a)












—
o
3
O
u
U
4-1
•C
Ol
5
~*








-
a
8
C
u
4J
-E
w
1



















ID
4J
O
1-

Ol
>
<
^_^
_Q



O
N
W>

U
o.
E
0)
(/)




•
o>
>
<




.—V
-Q

u
N
t/l
U
1
B
ID
VI





ai
<



n

* —
1)
N
\fi
tl
i
CO


















o>
>
0)


o
o
CsJ




m
CM



0
r%.
«s

o
rr\








r-.


IA

O"*
fl-
PN»


^O
DO


•O
o
en




^
j-



tA
^
ff

8
^
M










M
4-1
g

S
O
— i — -3-vo LA— r^cor--.
LAI — ao-7  \£)
vO —
ooooooooo
OLAOOLAOOOO
\D-T o^ror^r^tx^ moo
LA


CM\X> P^O>rn\O LANOvO
vO »— MO »— o oo \o r^ o
LA CM CO LA f\ O CO -* v*0
MO


CMCMLAO — — CMCTlLA
CMCMLA) — — — CT* LA LA
CMCMlArA — — LACMIA
MO —
\OLAO^OO —
— — ' — — O CM O !*> MO
^- MO —

•— •— onocorArACMr^
oocsf^cAfM^rvOfMr^
MOCi"imCMCMC«-\CO-3--3-
MO
CMOO-3-OOCMrMLACM —
csj^-rM-a-Jco-ajr--.
LAOfM — (Nil — ^MOCM-
r»


mMoJ"\*o\ommMOr^
LAO OCOOO LALAGOMO
r^rgm-fl'ojmoor^p^
MO
__
i^rAiA-a-MSLAr^-a-Mj
p^Mor^tM — OOLAP^
IA —

— CMOOO — OOLA-3-r~.
r--a- (j-icriroono r~«oo
trimrocMeMfMoo-a-*—
MO

mLAf-ivDMO-a1 r**p^i —
—  — — — IAPACM
— IA
I- In
V 4J
f L-
4-" II
IB C
tl
•— I/I —
u
L." 'i t!
VI V ID —
1) O J3 tl
4J 3 3 O •
v vi -a i- vi
w«o "O -^
in t i. - in cy
q a. o u ID o
J C •- — VI l-
I) L. 4-> — — VI
•D'O4)VI4->'OIDVI -
ou6.n!x;o*-'iTJ^:
OIDID— uou— vi
U-OQ.O-I— 3 4 o <
+i
o
o
+1
o
o
+!
O
o




+1
o
o



+1
o
o
+1
o
o


+ 1
o
o


+1
o
o




+1
o
o


+1
o
0

+1
§


+1
o
o

+1
o
o




+1
__
+1
o
o



^^
4->
C
u
u
U
JL

i/)
flJ
4J
&

13

-------
where X is the original percentage value of a component expressed
      as a decimal,

and   Y is the transformed value of X.

     After the data of Table 2 are similarly transformed and then the

sum of squares about the mean (SSM) is calculated for each component

within each weight group, the data in Table 3 are obtained.

     A commonly used measurement of relative precision is the coeffi-

cient of variation,  which is defined as the standard deviation (square

root of the variance) of a group of measurements divided by  its mean.

The result is then usually multiplied by 100 to convert to percent.

                           CV = (/V~ / Y )  100*

     To facilitate visual examination, the coefficients of variation

are summarized separately (Table k) .

     An examination  of the coefficients of variation of the three groups

indicates no discernible pattern.   For example, precision does not appear

to increase as the sample weight is increased.  The food waste and gar-

den waste components have relatively high coefficients of variation but

that of the paper products component is low.  Irregularly high coeffi-

cients include textiles in Group Ml  and wood in Group II, whereas the

coefficients of variation of the inerts component vary erratically.

     The Cochran C test5 may be employed to determine, in a rigorous

manner, (the Coefficient of Variation Table only provides insight)

whether the variances among sample weight groups can be pooled for a
    -'See Table 3 for definition of terms.

-------
       O4
       OQ
       i
       u.
              •z.
              o
a:
o
li I     gQ     (J

_J     O     Z

OQ            —

<     ^     M

h-     00     —
       co     a:
       —     o
       CO
       OL
       <->
              ac
              LU
             a


             §
            u)
                          -£,
                           en
-a-
£
OO
CM
O
o
LA
J^
O
O
On
O

•3"
O
LA
CM
-J
v-
O
0

en


-a-
o

o


_
•_
LA
O

CM
_
en




00
o
0


CM


vO
vO
1 —
O
o



_—
OO
CM
O















«j
t/l
II
X

1
"?
CM
00
O
o
o

CM
O
O

-
en
o
o
0

en

m
o
o

0


CM
o

o

LA
cn




en
CM
o
o


CM




o
o
o



CM

CM
O
















I/)
»

~
t>
t—
^°.
CM
en
0
o
o

s
o
o
0
CO
\O

O
j.
p;
LA
CO

0

en

r>-
LA
LA

O


\o
LA
LA
O

LT\
-a-
—



CM

O
O


CM




O
o
d



m

CM
o


















-,
1
o
Osl
^
,—
o
o

LA
O
o
00
Q

VO
O
0-N
0^
LA
0
o
o

en
MO
o

o

o


00
01
LA
0

00


Q.
3
O
L-
O1
ID
.E .
^-'

O.
E II
8 z
01
10 en

i_ ..
O c
U- ID
-^ 2:
•o
O) ol
I J=
O "
11



*— -* I/J
C (D
O 3
— 0"
fD
> 4)

(/l
^ *^
o o

c c
o
u- •—
o <->

c £
ra E
i^
« ji

.1^ .12


Wl
00
>
$
•K


















































J
1



II
Q

.«
Q
•o

4)
l_
**•
14-
o

VI

11
L.
O1
V
•o
£

.11


a









































c
0
4-1
ID
>
U

o
•a o
-5 "~
U —^»

t;1?
u
4-*
I/I O


> c
4-t
U- as
0 —
Z ID
to >

II O
=» 4-1
c
.. u
41 ._
U U
c •-
ia u-
— 14-
 u
V 0

.— .!2


,j






                                                  15

-------
         TABLE 4




COEFFICIENTS OF VARIATION




(EXPRESSED AT PERCENTAGES)
Components
Food waste
Garden waste
Paper products
Plastic, rubber, leather
Text i les
Wood
Metals
Glass and ceramics
Ash, rocks, dirt (inerts)
GROUP 1
31.2
1*8.9
0.5
15.4
19.5
14.5
13.8
10.9
37.8
GROUP 1 1
23-5
51.0
6.6
27-0
11.9
77.4
9-9
15.6
189-3
GROUP 1 1 1
31.4
26.3
3.5
19-9
57.9
29.6
21 .0
13.7
7-5
          16

-------
given component.  The test compares the sum of a group of variances

with the largest member of the group.  Results of the test  indicate

that, at the 5 percent risk level, pooling of variances across groups

is indeed justified provided the "questionable deviates," represented

by the textile component in Group  III and the wood component  in Group

II, are removed.

     Additional use of the Cochran C test indicates that pooling

variances among components is also justifiable.  Thus, a single pooled

estimate of variance, which yields an estimated standard deviation,

s, for all  transformed observations of 0.1413 at 72 degrees of freedom,*

is obtained from Table 3-

     The conclusion to this portion of the analysis is that there  is

no difference  in precision among the three groups.  Accordingly, a

"small" (20G-lb) sample  is as adequate as a larger one.  The  desirabil-

ity of taking small samples, supported by the preceding statistical

analysis, is reinforced by the fact that they are much easier to handle.

Past experience has shown that separating "small" samples is  simpler

and more accurate from a human standpoint than separating "large" ones.


             Statistical Technique to Determine Sample Size


     To determine the number of samples required for composition anal-

ysis, the following normal  approximation is adequate.
    *As percentages, there are only eight independent observations per
sample because the ninth component is predictable after the percent
values for the first eight have been determined.  The transformed values,
however, are not subject to such predictability and thus yield nine in-
dependent observations per sample.  All components from the samples in
the three groups are included in this estimate, as this best reflects
conditions to be found in practice during a 4-day sampling period.

                                  17

-------
     where   z is the standard normal deviate for the confidence
             level  des i red,

             s is the estimated standard deviation (transformed
             bas is) ,

             6 is the sensitivity (transformed basis), defined
             below,

     and     n is the number of samples required.


     At 90 percent confidence, z equals 1.6^5; s equals 0.1^13, as

just noted.  The sensitivity, transformed basis, is determined as follows

transform X and either X + A or X - A by the previously mentioned arcsin

transformation (specific values shown in Table 5), where X is the ex-

pected percentage of the component in question and A  is the desired

precision for the percentage to be estimated (see example for further

clarification of A).   The choice of sign for X ± A is positive if X

is less than 0.50 and negative if X is greater than 0.50.  Calculate

the sensitivity,

             I           i—             i	  I
         6 = | 2 arcsin /X -  2 arcsin /X ± A  j


and compute n (number of samples required) by the formula given.

     Example.  Suppose one wishes to estimate the percentage of metals

in raw solid waste to within two (2) percentage units with a confi-

dence of 90 percent.   Metals are expected to be found at a concentration

of approximately 9 percent.   The first step  is to convert the average

-------
       TABLE 5




ARCS IN TRANSFORMATION




   Y = 2 arcsin /X
X
.001
.002
.003
.004
.005
.006
.007
.008
.009
.010
.011
.012
.013
.014
.015
.016
.017
.018
.019
.020
.021
.022
.023
.024
.025
.026
.027
.028
.029
.030
.031
.032
.033
.034
.035
.036
.037
.038
.039
.040
Y
.0633
.0895
.1096
.1266
.1415
.1551
.1675
.1791
.1900
.2003
.2101
.2195
.2285
.2372
.2456
.2537
.2615
.2691
.2766
.2838
.2909
.2978
.3045
.3111
.3176
.3239
.3301
.3363
.3423
.3482
.3540
.3597
.3654
.3709
.3764
.3818
.3871
.3924
.3976
.4027
X
.041
.042
.043
.044
.045
.046
.047
.048
.049
.050
.06
.07
.08
M
.10
-n
.12
.13
.14
.15
.16
.17
.18
.19
.20
.21
.22
.23
.24
.25
.26
.27
.28
.29
.30
• 31
• 32
.33
.34
.35
Y
.4078
.4128
.4178
.4227
.4275
.4323
.4371
.4418
.4464
.4510
.4949
.5355
.5735
*'«
.6435
**|6!
.7075
.7377
.7670
.7954
.8230
.8500
.8763
.9021
.9273
.9521
.9764
.0004
.0239
.0472
.0701
.0928
.1152
.1374
.1593
.1810
.2025
.2239
.2451
.2661
X
.36
.37
.38
.39
.40
.41
.42
.43
.44
.45
.46
.47
.48
.49
.50
.51
.52
.53
.54
.55
.56
.57
.58
.59
.60
.61
.62
.63
.64
.65
.66
.67
.68
.69
.70
.71
.72
.73
.74
.75
Y
1 .2870
.3078
1.3284
1.3490
1.3694
1.3898
1 .4101
1.4303
.4505
1.4706
1 .4907
1.5108
1.5308
1.5508
1.5708
1.5908
.6108
.6308
.6509
.6710
.6911
.7113
.7315
1.7518
1.7722
1.7926
1 .8132
1.8338
.8546
.8755
.8965
.9177
.9391
.9606
.9823
2.0042
2.0264
2.0488
2.0715
2.0944
X
.76
.77
.78
.79
.80
.81
.82
.83
.84
.85
.86
.87
.88
.89
.90
.91
.92
.93
.94
.95
.951
.952
.953
.954
.955
.956
.957
.958
.959
.960
.961
.962
-963
.964
.965
.966
.967
.968
.969
.970
Y
2.1177
2.1412
2.1652
2.1895
2.2143
2.2395
2.2653
2.2916
2.3186
2.3462
2.3746
2.4039
2.4341
2.4655
2.4981
2.5322
2.5681
2.6062
2.6467
2.6906
2.6952
2.6998
2.7045
2.7093
2.7141
2.7189
2.7238
2.7288
2.7338
2.7389
2.7440
2.7492
2.7545
2.7598
2.7652
2.7707
2.7762
2.7819
2.7876
2.7934
X
• 971
.972
.973
• 974
.975
.976
.977
.978
.979
.980
.981
.982
.983
.984
.985
.986
.987
.988
.989
.990
.991
.992
.993
.994
.995
.996
.997
.998
.999











Y
2.7993
2.8053
2.8115
2.8177
2.8240
2.8305
2.8371
2.8438
2.8507
2.8578
2.8650
2.8725
2.8801
2.8879
2.8960
2.9044
2.9131
2.9221
2.9315
2.9413
2.9516
2.9625
2.9741
2.9865
3.0001
3.0150
3.0320
3.0521
3.0783











          19

-------
percentage and the acceptable precision to decimals, i.e., 0.09 and

0.02, respectively.  The corresponding values of Y  (the transformed

quantities) obtained from Table 5 (X equals 0.09 and X + A equals 0.11,

since X is less than 0.50)  are 0.6094 and 0.6761.  Thus, & = 0.0667,

and

                                (0.1413)          ..
                                                 ']k
or n is approximately equal to 12 samples.

     By performing a similar computation for each of the nine compo-

nents and by using the expected percentages based upon the averages

of the Group III samples, the following number of samples required

for each component are obtained (assuming 90 percent confidence and

a reasonable error of no more than 2 percent).

     Component                      Number of samples size

     1 . Food waste                            9
     2. Garden waste                          6
     3. Paper products                       20
     ^. Plastic, rubber, leather              8
     5. Text! les                             10
     6. Wood                                 10
     7. Metals                               11
     8. Glass and ceramics                   11
     9. Ash, rocks, dirt (inerts)            12

     If all components are considered equally important to the total

analysis and if a conservative approach is taken, a maximum number of

20 samples  is needed to analyze the composition of solid waste.  With

the exception of the paper component, however, 12 samples should prove
                                  20

-------
adequate."  Further, within the confines and limitations of the study,

the preceding analysis suggests that 12 of the smaller (200-lb) sam-

ples will yield valid and reliable results for composition evaluation.
    "This conclusion is, of course, based upon an assumption that at
90 percent confidence,  reasonable error should be limited to no more
than 2 percentage units.  If this criterion were relaxed, the necessary
sample size would be reduced accordingly.
                                 21

-------

-------
                        SUMMARY AND CONCLUSIONS






      Important to  the study of problems associated with solid waste




management  is the  definition, analysis, and evaluation of the generated




waste.  Such materials entering a solid waste management stream, re-




quire storage, collection, handling and processing, and ultimate dis-




posal or salvage and reuse.  Developing a reliable and standard sam-




pling procedure to determine the composition of solid waste materials




generated from any of a multitude of sources is of primary  importance.




      It is  undesirable and impracticable to separate, measure, weigh,




and analyze all the solid waste generated at specified sources or




disposed of at particular facilities.  Thus, to balance the resources




available with the need for accurate results, appropriate statistical




analysis holds promise of yielding information regarding optimum weights




and number  of samples to be taken.  By initiating a planned sample sur-




vey,  the composition of solid waste can be determined and evaluated




within acceptable  statistical limits.




     To draw conclusions and to make recommendations based upon the




preceding analysis, the study analyzed one incinerator operation and,




more specifically, its solid waste composition for a 't-day period of




time.  A more thorough and complete analysis would also consider monthly,




seasonal,  yearly,  and cyclical variations.  Note also that solid waste




must be properly mixed to yield adequate samples from a given supply
                                 23

-------
of waste materials; and further, that if the criteria to classify solid




waste differ from those applied in selecting the nine categories used




in this analysis, the results may be influenced significantly.




     With these facts in mind, the study of this incinerator indicates




that there is no significant variance among sample weight groups.  Con-




sequently, smaller (200-lb) samples will yield valid results in deter-




mining the composition of a given supply of solid waste.  Data from




later studies, which are presently being evaluated and analyzed, appear




to confirm and verify the results of this analysis.  Indeed, it may




be found that samples somewhat smaller than 200 Ib will provide accep-




table results.




     It is encouraging to note that, upon the basis of this study,




smaller samples than previously thought adequate hold promise of pro-




viding reasonable and reliable information concerning the composition




of solid waste.

-------
                               REFERENCES
1.   Brown, R. R.  Explanation in social  science.  Chicago,  Aldine
         Publishing Company, 1963-  198 p.

2.   The Solid Waste Disposal  Act; Title  II  of Public Law 89-272,
         89th Cong.  S.306, October 20, 1965.  Washington, U.S.
         Government Printing Office, 1966.   [5 p.]

3.   Bell, J. M.   Development  of methods  of  sampling and analyzing
         municipal  refuse 1957~1962.  Unpublished paper.  Purdue
         University Library.  20 p.

k.   Cochran, W.  G.  Sampling  techniques.   2d ed.  New York, John
         Wiley and  Sons,  Inc., 196^t.  p.  71.

5.   Winer, B. J.  Statistical principles  in  experimental  design.
         New York,  McGraw-Hill Book Company  Inc., 1962.   p.  9^-95.
                                  25

-------

-------