Module4:det

erminati
onofsampl esize–sampl i
ngerr
ors-
datacoll
ect
ionanddata
edi
ting-codi
ng-cat egor
izat
ion-handlingunsat
isf
act
oryresponsesandmi ssi
ng
val
ues-questi
onnair
econst r
ucti
on-t ypeofquesti
ons-goodnessoff i
tofdata-
rel
i
abil
it
yandconsi
stency-Cronbach’
sAl pha

Sampl
ingEr
ror
Asampl
i
nger
rori
sast
ati
sti
caler
rort
hatoccur
swhenananal
ystdoesnotsel
ecta
sampl
ethatr
epr
esent
stheent
ir
epopul
ati
onofdat
aandt
her
esul
tsf
oundi
nthe
sampl
e do notr
epr
esentt
he r
esul
tst
hatwoul
d be obt
ained f
rom t
he ent
ir
e
popul
ati
on. Sampl
i
ng i
s an anal
ysi
s per
for
med by sel
ect
ing a number of
obser
vat
ionsf
rom al
argerpopul
ati
on,andt
hesel
ect
ioncanpr
oducebot
hsampl
i
ng
er
ror
sandnon-
sampl
i
nger
ror
s.

 Sampl
i
nger
rori
sast
ati
sti
caler
rort
hatoccur
swhenananal
ystdoesnot
sel
ectasampl
ethatr
epr
esent
stheent
ir
epopul
ati
onofdat
a.

 Ther
esul
tsf
oundi
nthesampl
ethusdonotr
epr
esentt
her
esul
tst
hatwoul
d
beobt
ainedf
rom t
heent
ir
epopul
ati
on.

 Sampl
i
ng er
rorcan be r
educed byr
andomi
zi
ng sampl
e sel
ect
ion and/
or
i
ncr
easi
ngt
henumberofobser
vat
ions.

Ty
peIEr
ror
Aty
pe Ier
roroccur
swhent
he 
nul
lhy
pot
hesi
s i
str
ue,
buti
srej
ect
edf
alse 
byt
he
t
est
ing.
AtypeIerror
,orfal
sepositi
ve,isasser
ti
ngsomethingast
ruewheniti
sactuall
y
fal
se.
 Thisfal
sepositi
veerrorisbasi
call
ya“f
alsealar
m”–aresul
tthati
ndicat
esa
giv
enconditi
onhasbeenf ulfi
l
ledwhenitact
uall
yhasnotbeenf
ulf
il
led(i
.e.
,
err
oneouslyaposit
iveresul
thasbeenassumed) .

Ty
peIer
ror
sar
eal
socal
l
ed:
Producer
’sr
isk,
Fal
seal
arm,  
err
or.
Ty
peI
IEr
ror
Atype I
Ier
roroccurswhent
henul
lhy
pot
hesi
sisf
alse,buter
roneousl
yfai
l
stobe
r
eject
edbythetesti
ng.
AtypeIIer
ror
,orf
alsenegat
ive,
iswhereat
estresul
tindi
catest
hatacondi
ti
onfai
led,
whi
leitact
uall
ywassuccessf
ul. 
 ATypeI
Ier
roriscommi t
tedwhenwefail
tobeli
eve
atr
ueconditi
on.
Ty
peI
Ier
ror
sar
eal
socal
l
ed:
Consumer
’sr
isk,
Misdet
ect
ion,  
err
or

Tabul
arr
epr
esent
ati
onofTy
peIEr
rorandTy
peI
IEr
ror
Nul
lHy
pot
hesi
sist
rue Nul
lhy
pot
hesi
sisf
alse
Rej
ectnul
lhy
pot
hesi
s TypeIEr
ror
Fal
se Cor
rectOut
comeTr
uePosi
ti
ve
Posi
ti
ve
Fail
tor ej
ectnul
l Corr
ectout
comeTr
ue Ty
peI
IEr
ror
Fal
seNegat
ive
hypothesi
s Negati
ve

Exampl
es:
Ty
peIEr
ror
Let’
suseashepherdandwol fexample. Let’
ssayt hatournullhypot
hesi
si st
hat
ther
eis“nowolfpresent.”
 At ypeIerror(orfalsepositive)woul
dbe“ cr
yingwolf
”
whentherei
snowol fpr esent
. Thati
s,the actualconditi
on wasthatt
herewasno
wolfpr
esent
;however,theshepherdwr onglyindicat
edt herewasawol fpr
esentby
call
i
ng“Wolf
!Wolf
!”
 Thisi satypeIerr
ororf al
seposi t
iveerror
.
Ty
peI
IEr
ror
Continui
ngourshepher dandwol fexampl e. Again,ournullhypot hesisi
sthatthere
i
s“ nowol fpresent
.” AtypeIIerr
or( orfalsenegative)wouldbedoi ngnothing(not
“cry
ingwol f
”)whent her
eisactuall
yawol fpresent.
 Thatis,the actualsi
tuati
on was
thatther
ewasawol fpresent
;howev er,t
heshepher dwr onglyindicatedtherewasno
wolfpresentandcont i
nuedtoplayCandyCr ushonhi siPhone. Thi sisatypeIIerr
or
orfalsenegativ
eerror.

Nul
lHy
pot
hesi
s Ty
peIEr
ror/Fal
sePosi
ti
ve TypeI
IError/Fal
se
Negat
ive
Wol
fisnotpr
esent Shepherdthi
nkswol
fis Shepherdthinkswolfi
s
present(
shepher
dcri
es NOTpr esent(shepher
d
wolf)whennowolfi
s doesnothing)whenawolf
actual
lypr
esent i
sactuallypresent

CostAssessment Costs( act

ual costsplus Replacementcostfort
he
shepher dcredibi
li
ty) sheepeat enbyt
hewolf,
associatedwi thscrambli
ng andr epl
acementcostfor
thetownsf olktokil
l t
henon hir
inganewshepher d
-exi
sti
ngwol f
Dat
aPr
ocessi
ng
 Afterrecol
lect
ingf i
l
ledquesti
onnair
eswi t
ht her equir
eddata,ithasbeen
processed,thesurveyquesti
onnair
esi [Link] eralsteps
i
nv ol
vedinanal y
sis,suchaschecking,edi
ting,coding,andtranscri
bing,as
wellasdesignati
nganyspecialoreccent
ri
ctreatment sofdat
aaf oretheyare
analyzed.
 Aftervi
ewingtheQuest i
onSummar ies,Onecancr eaterulestoanswermore
specif
icquesti
[Link]
lter
,Compar e,andShowr uleshel
pstofocus
i
n on specifi
c subsets ofthe data,so thatanalysis ofresul
tsis most
meaningful
.Itisimportantt
okeepi nmi ndt hatdataanalysi
sisaniter
ati
ve
[Link]
enr esul
tsf r
om onet y
peofanal ysi
smayy i
eldnew quest
ions,
resul
ti
nginanotherroundofdataanalysi
s
 Thereareseveralprocedur
eswecanusetoanal
yzeourneedsassessment
[Link]
sesmet
hodsar
e:
[Link]
ipt
iveSt
ati
sti
cs:
Descri
pti
ve stat
ist
ics descri
be and summari
ze t
he quant
it
ati
ve
i
nformati
on f
rom y oursurveyaboutyoursampl
e.I
tincl
udessuch
summar ymeasuresas:
• Mean(
theav
eragev
alue)
• Median(t
heval
ueatwhi
chhal
foft
hev
aluesar
eabov
e,and
hal
farebel
ow)
• Frequenci
es(theper
cent
ageofr
esponsest
hatf
alli
ntoaset
ofcategor
ies.
)
[Link]
oss-
tabul
ati
ons:
Simpl
ecross-t
abul
ati
onsal l
ow y
outocomparetheresul
tsoft
womut ual
l
y
excl
usi
vegroupsofrespondent
[Link]
egor
ieswe
cancompareasmanygr oupsaswel
ike.

Dat
aChecki
ng
Datacheckinginvol
v eseli
minat
ingunacceptabl
equesti
onnai
[Link]
ear eseveral
reasonswhyaquest ionnai
remaybeunaccept abl
eforuseinastudyli
keitmightbe
i
ncompl et
e,inst
ruct
ionsnotfoll
owed,l
it
tl
ev ari
ance,
missi
ngpages,pastcutof
fdate
orrespondentnotquali
fi
ed.
Dat
aEdi
ti
ng
Thesecondst epi stoedi tthedat abyex ami ni
ngt hequesti
onnai
retoidenti
fy
answersunr eadabl
e,Incomplete,incompatible,unsati
sfact
ory,equiv
ocalanswers
thatwil
lbescr appedinsteadoff i
ll
ingthemi ssingv al
uetoreducetotheminimum
therepl
icat
ionbi asthusstrengtheningtheprecisionandt heaccur
acyofthedata.
Dataediti
ngcanbeper f
ormedmanual ly,witht heassist
anceofacomput erora
combinati
onofbot h.
Dat
aCodi
ng
 Once we completed oursurvey checki
ng and edi
ti
ng;now we need to
minimizebaddat
aasbestwecoul dsowemayneedt ocodesomeresponses
i
nor dertoenabl
eanalysi
[Link]
gni
ngnumer i
c
oralphainf
ormat
iontoquesti
onresponsesthatdonotordi
nari
l
yret
urntothe
resear
cherint
hatfor
mat .
 Codingtypicall
yassignsanumbert oanswer sthatdonotal
readyhav ethem
so thatstatist
icaltechniquescanbeappl [Link] hatusually
requir
e coding ar e often demogr aphic – age,gender,mar i
talst at
us,
[Link] heori
ginaldatamayal r
eadybenumer i
csuchas
age,butcodingwi llpr
ovideforanaggr egat
ionintomuchsmallandt her
efore
mor eusef
ul categori
es.
 Inresearch,codingofquali
tat
ivedatawithhel
pofLi kertscal
ehasbeendone
asifpartici
panthaspercepti
onincompl et
eresemblancewi thwhateverasked
i
tdenot edas5,i fr
espondentismostlyagreeitdenotedas4,whent hereis
notanyper cepti
onorcer t
ainideaitdenotedas3i frespondentisrareagree
aboutthei nfor
mat i
onaskeddenotedas2andwhenv eryrareagr
eedenot ed
as1.
 SPSShasbeenusedf
orbot
hdat
acodi
ngaswel
lasDat
atr
anscr
ibi
ng.

Dat
aTr
anscr
ibi
ng
Datarecordedf r
om v er
balint
eract
ionwithrespondentstobet
ranscri
bethatisto
convertspeechsoundi ntowordsasaccur at
elyaspossibl
[Link]
anscri
binginv
olves
l
ist
eningt or ecordeddataandt ypi
ngoutexact lywhatwassai [Link] i
mes
tr
anscribeasect ionoft extintoelect
ronicform ei
therbyscanningitwithan
opti
calcharacterrecognit
ionprogr
am, orbydirect
lyt
ypingi
toncomput er
.
Dat
atr
anscr
ibi
ngEssent
ial
s
Consi
stency
Consi
stencyi
ntranscri
bingmakesworkunderstandable,andmakesi tpossi
ble
totur
ni tint
o displayabl
[Link] you t
ranscr
ibe features ofthe l
ayoutand
for
matti
ngofyourtextshouldr
emai
nconsist
entthroughoutyoursubmissi
on
Compl
eteness
Compl etenessmeanst hatwehavetr
anscr
ibedallt
herel
evantsecti
onsofthe
textthatwer egi
ven,andcheckedt
oensur
et hatwehav
enotmi ssedanypages,
paragraphs,sent
ences,l
ines,
orwor
ds.

Cor
rect
ness
.Correct
nessmeanst hatwehaverespectedthewordi
ng,spel
l
ing,
capit
ali
zati
on,
i
tali
cs,bol
ding,punct
uati
on,
[Link]
heor i
ginalt
ext
,andrecr
eatedi
taccurat
elyin
yoursubmission.

Dat
aAnal
ysi
s

Analy
sisofDataisaprocessofinspecti
ng,cleaning,t
ransf
orming,andmodeling
datawit
ht hegoalofdi
scover
ingusefuli
nformation,suggest
ingconclusi
ons,
and
support
ingdeci
sion-
making.I
0tincl
udestwot ypesofanalysi
s:

[Link]
ipt
iveAnal
ysi
s
 Thetransfor
mationofr aw datai
nto aform t
hatmaket hem easyto
under
standandinterpret
;rearr
angi
ngthecommandandt hemanipul
ati
on
ofdat
at ogener
ateinformati
oniscall
edDescr
ipt
iveAnal
ysis.
 Thedescri
pti
veanalysisprovi
desaconcisesummar yoft
hedat [Link]
summarizedatadigi
tall
yorgr aphi
cal
ly
.Thedescri
pti
veanalysi
st akesthe
for
m ofdescri
pti
vest ati
sti
[Link]
pti
vest ati
sti
csis
tosummar i
zeasetofquant i
tati
vedat
[Link]
chersusethesest ati
sti
cs
todescr
ibeorchar
act er
izethesamplestudi
ed.
 Thedescri
ptiv
eanal ysi
sisthetr
ansformati
onofraw dat
aint
ocompl ete
i
nfor
[Link] susedt odeter
minet hemai ntr
endsofthev ar
iabl
es.
Ordi
nar
yt echniquesusedi nthestat
isti
csincl
udethemeasurementof
cent
ral
tendency ,av
erageandmedian.
 Otherdescri
pti
vest at
isti
csaremeasuresofthevari
anceandstandard
devi
ati
[Link] i
sualtool
sofDescr
ipt
iveAnal
ysi
s,such
asbargraphs,Pi
eChar tsandLi
negraphs.

2.I
nfer
ent
ial
Anal
ysi
s
 Inf
erent
ialstat
ist
ics att
emptst orespond tothe quest
ions on the
populat
ionsandsampl esthathaveneverbeentest
edinthepar ti
cul
ar
experi
[Link]
erenti
alanalysi
si s anxi
ous t
o make predict
ions or
 i
nferencesonapopul
ationfrom obser
vat
ionsandanal
yzeswhichmeans
thatwecantaket
heresultsoftheanal
ysi
sofagivensampleandthenwe
couldgener
ali
zet
othelargerpopul
ati
on.
 Todothis,
iti
si mperati
vethatt
hesampl eisrepr
esent
ati
veoft
hegr
oupto
whi
chitgenerali
[Link]
ysi
sisrequir
edonlywhenasamplei
spul
led
byarandom procedureandtheresponserateisver
yhi
gh.
 Theref
ore,thistypeofanal ysi
si snotappr opri
atewhent hemet hodsof
Non-probabi
li
tyselecti
[Link] heInfer
enti
alanalysi
si sa
corner
stone ofr esearch,because i tis generall
y dif
fi
cult
,and of ten
i
mpossi bl
e,toprobeallmember sofapopulationortoobserveeachev ent
.
Inst
ead,theresearchersaret r
yingtoobtainar epr
esentati
vesampl e,and
useitasabasi sformor egeneralconcl
usions.

Cr
onbach’
sAl
pha
Cronbach’
sal pha,α (or 
coeff
ici
ental pha)
,dev el
oped byLeeCr onbach in1951,
measuresreliabil
it
y,or 
int
ernalconsist
ency.“Reli
abil
i
ty”ishowwel latestmeasures
[Link] e,acompanymi ghtgiveaj obsatisfact
ionsurveytothei
r
[Link] ghreli
abil
it
ymeansi tmeasuresj obsat i
sfact
ion,whilelow r
eli
abil
it
y
meansi tmeasur essomethingelse(orpossiblynothi
ngatal l
).
Cronbach’salphateststoseei fmul ti
ple-
question 
Liker
tscale 
sur
veysar ereli
able.
Thesequestionsmeasur e l
atentvari
ables 
—hi ddenorunobser v
abl
ev ari
abl
esl i
ke:a
person’
s conscienti
ousness,neur osis [Link] ar ev ery di
ff
icultto
measurei nrealli
[Link]’salphawi l
ltel
ly ouifthetestyouhav edesignedi s
accurat
elymeasur i
ngthev ar
iableofinter
est.
Cr
onbach’
sAl
phaFor
mul
a
Thef
ormul
aforCr
onbach’
sal
phai
s:

Wher
e:
N=t
henumberofi
tems.
c̄ = 
aver
age 
cov
ari
ance 
bet
weeni
tem-
pai
rs. v
̄ =av
erage 
var
iance.
Accept
anceofaCr
onbachScor
e