A ei
6.1. Introduction to Sampling
Sn sample a small piece of cheese b
Shoppers oft ore pure
‘ene piece what the lagerchunk wil asic like Achsmis does
sample of alcohol frm a sill, determines that is 90 prof
fol in the stills 90 proof the chemist tests ll the alcohol or th
cheese, there will be none to sel. Testing al of the product often desto
Grou, To determine the characteristics of the whole we have to sample only
‘Suppose thats the personnel director of a large bank. you need io writes cep
scribing all the employees who have voluntarily left the
would have a difficult tsk locating ll these thousands of people. T
cessible a8 a ex0up—many have died, moved from the community, left the
Guited a new rame by mariage. How do you writ the report? The best
representative sample and interview them inorder to generalize about the entire group.
Time is also a faclor when managers nee information quickly in order to adjust a
‘eration orchange a poley, Consider an automatic machine that sons thousands of pieces of
mail daily, Why wai for an entire day's output to check whether the machine is work!
yoo es
ion character by the postal se
accurately (whether the popul
vice)? Instead. sainples can be taken at specific intervals, and if necessary
be adjusted night away
Ceowscrsamle ‘Sometimes its possible and practical to examine every person or item in the population
‘ae wish to describe. We call thisa complete enumeration, or census snp ee
itis not possible to count or measure every item pul
fampesslppbton sndsemls Statisticians use the word population to refer not onl to people but toa items that have
sare all the cheese
Deen chosen for stu Inthe eases we have jut mentioned, the pop
the chunk, all the whiskey in the vat al the employees ofthe large bunk
in the last 10 years and all mail sorted by the automatic machine since
Check. Statisticians use the word sample to describe a portion chosen from the population.
voluntarily Tet
Statistics and Parameters
Fcond sii ‘Mathematically. we can describe samples and pop:
anépoones mean, median, mode, and standard deviation, which we i
these terms describe the characteristics of a sample, they are called statistics. When they
describe the characteristics of a population. they are called parameters. A statistic Is &
ci characteristic ofa sample; a parameter isa characteristic of a population,
lations by using measures such as the
joduced in Chapter 3. When
296 — chapteré sampling and Sampling Distributions.gw meae
Beas erciennine ore
peas,
shed. At its
\e president of PSE,
snplications
PE er ha on ih paises tsiner
al ge eee Tae creeat ese ean
pars eievall sasionte bo ities
ipulbntotdare whan hey appro ofthe eal Shy pce
ae ere ere
eR are aes
os jennion,
id Jean oward the upcoming union vote, met with some difficulty after re
PEN oe treater
ting her findings
omen and from the epinning datay it was clear (oro Jean though) hak
samp joes were favoring a unionized shop. Jean's eport was shrugged
toe ettk the comment, “This is no g004. Nobody can make statements about
etree sentiments when she talks to only a itl over 15 percent of our
srr ofees. Everyone knows you have to check SO percent to have any Wea
ero hpe outcome of the union vote will be, We didn’t hire you to make
guesses." Is there any defense for Jean's position?
gg 67, Aconsumer protection organization is conducting a census of people who.
a were injured by a particular brand of space heater. Each victim js asked ques-
Tons about the behavior of the heater just before its malfunction; this infor-
ilable only from the victim, because the heater in
tself upon malfunction. Early in the census, itis
mation generally is availabl
question tends to incinerate itself
sees ered that several of the victims were elderly and have died. \s any cen-
je? Explain.
sus of the victims now possibl
5.2 Random Sampling
Ina random or probability sample, we know what the chances are that an element of the
ca willor will not be included in the sample. As a result, we can assess objectively
the estimates of the population characteristics that result from our sample; that is, we can
| Cup Siping and sampling Disibtonspening te ane
vnmpbcerot
anininte population
ed
et
cr)
pened
Ped
Cetra
rs
Simple random sampling
Systematic sampli
Stratified sampling,
Cluster sampling,
Simple Random Sampling
Jecis samples by methods that sllow each possible somple
Simple random sampling s\
aaaee reaual probability of being picked and each Item inthe entire popula
ded tn the sample, We ean iusrae These te
‘infos of four students in a semitar and we
nen Tale 6-2 Wisirtes
fan equal chance of being in iwernente
‘vith an example. Suppose we have a popul
Want samples of two students ata time for interviewing purpe#
Uf the possible combinations of samples of two students in = Po
or seaity of each sample being picked and the probability hat each seen
a sample
ATP: ape illustrated in Table 6-2 uses ait population of tow mens Oy FE
Feet sc pepuasion has sated rite sie hat isto sy ek a Se
re mean ne Patou many ers here aren he poplin Cosas we SE
Tee ea seeing te sudent we shall soon exhaust our stall paises Tes
Sa ee fre ocerce (oat a CS relees Be UR eee a
eae eae ceo eee
‘could appear twice in the sample
ly to help vs think about sampling from an infiite pops
‘We have used this example onl
in which its theoretically impossible to ob-
lation, An infinite population is 2 population i
serve all the elements. Although many ‘populations appear to be exceedingly large, no tly
se population of physical objects actully exists, Aft 2 ‘given unlimited resources
saryiige, we could endmeate ay Finite popolaion een He Be ‘of sand on the beaches
are ith America, As practical mater. then, we wills he term infinite population when
of Nom king about a population that could not be enumerated in ‘a reasonable period of
‘students A 8, and 0
Poo samples of two people: AB, AC AD, BG 80.9
Frebabiyof raving is same of Two pele must
ia) ="
jase ao spss
clzwo peel)
ple ust De
probability ofthis student in thesaoe tasers _wosnes| \ oe
o
ton many closely spaced
sampling The easiest way to select a sample randomly is 6956
dene an be generated either by a computer PrOBFTHME LOSE
lon numbers. whieh should properly be called a rable of
7 Hove to Do Rando!
down nurnbers: The
ble numbers or by a table of ran
random digits
reece illustrates ponton of such able. Here we have, 1.150 random digits in SS
of lO gio, These numbers have Been generated by a completely random proces. Te
probability that any one digi from O through 9 wil appears the sane as that for any other
seit andthe probability of one sequence of digits occurring i he same that for any
ther sequence of the same length.
“To see how to use this table. SUPP*
wish to interview a randomly chosen sarny
sample by assigning cvery employee # nue
picking 2 systematic method of selecting tw
following:
Jose that we have 100 employees in acompany and
ple of 10. We could get such a random
ber of 00 to 99, consulting Table 6-3. and
jo-digit numbers. In this ease, let's do the
plrandom cigs
Sampling and Sampling DistributionsGo from the top to the bottom of tha columns beginning with the left-hand col
sian nt ony the it wo dis ech fow Notis tatuOr a
4, rin Ns mati wou be he ond he hind At, and 80 07
-_Htwereh them of hat column on herp nd ae ort of oUF
cred 10 twoetight nurbers of 9 and under, we can go beck othe bee
i at hand column) and start reading, the third and fourth
digits of each number, These would begin 81, 28, and 12
Another way to select our employees would be 10 write the name of each one om sip
Seen apm he ac een Arig em Omori oe cul 1
Slips at random. This method works well with a small group of people but presents peob
Jems ithe people inthe poputetion number in the thousands, There isthe added proiem
for example, when capsules were drawn from a bow! to determine by bir the onder
eer etter aredsrvien, Decens inhays wpe more fen ha
pba weld abe oes As wna Decertber capes sd been
placed in the bow! lst, and the eapsules had not been mixed proper on December €39
Re crceuaawe oe
Systematic Sampling
In systematic sampling, elements are selected from the population atau
is measured in time, order, oF space. If we wanted to interview every t
‘acollege campus. we would choose a random starting point in the fin
dent directory’ and then pick every twentieth name thereafter
Systematic sampling differs from simple random sampling in that cach clemers
equal chance of being selected but each sample does not have an equal chancs of Be
ected, This would have been the case if, in our earlier example, we had as
between 00 and 99 1 our employees and then had begun to choose a sample of
ing every temth number beginning 1, 11,21, 31, and so fon. Employess =
‘and 5 would have had no chance of being selected together
process. Suppose
Pimple 100 households every Monday. Chances are high Wht o¥f sample would not be
ar ative, because Monday"s trash would very likely include the Sunde nop
Tre ampunt of waste would be biased opward by our choice ofthis samplnE Pros
Systematic sampling has advantages, (00, howeves, ESRD though system
soay be inappropriate when the elements lena sequential Dem this method may
Tess time and sometimes results in low’
Stratified Sampling
require
awe use one of two approaches. Bither w
€ propo
called strata.
stratum a specified number of ele
the population as a whole of we J
give weight 10 the FesuTIs ae
ther approach, stratified sampling guarantees thst
‘chance of being selected.
ments corresponding (0 th
ments from each strat
Fs proportion of total populti
every element in the pop
6.2 Random Sami
on, With
putation has &
Te systematic sampling, there isthe problem of introducing an error ie Os SANE
svc were sampling paper waste produced by households, ard we decides 9
er. Thus
fer costs than the simple random-sample method.
“To use stratified sampling, we divide the population 19 relatively homogencous 20U2%
hi fe select at random from ea
ans
301compton snd
iste wosg
npn se dom sang
02 chapter 6 Sampling
je akeady divided
aie pooaion .
ee 8
at Sug tT ysl na
tof charac
us iat
the four a8
rors his ate roups and give
Gre divided io FOO
To owt hoe many
‘tthe pepain. he ou
erento he samples cor
sere of a stati ar
sya of aie saps
ecuately refet charac he pop
‘Siner kind of ame
stake aad
ing 1 He P=
sare ox gatets that oHP-T
ore self-C
properly designed. they
exer
hen they a
Se ele ey ee chee ae
sstes oft
Cluster Sampling
Incluter sampling, w=
dom sample of these closers
‘tite population asa whole
pling tbe average number of
Ey map to divide hee
fe) for inleniewing
wwelhdesigned cise 5
Shy less eos than that of Sr
yes oe an acute sarmpling; he poelation is SA"
spon We ue said sampling when e2ch SO SSE variation withi
Nea Taser sampling inthe cops
imo groups clters, and then sect
that these individaat
nica isalterping
meena in a large city He
retoose acertain number of IGE
ose Hocks would be interviewed. A 4
fore precise sample at consier Basic ‘
it the popula
a masket res
television sts Pt
nro blocks and th
id in exch oft
ring procedre ean posuee 3
ple random sampling
determine by s-
vy eould use 3
Soop ide saan betacen he groups. Weuse cle SMPTE
saan nn winch ep ble TOURS a