Cour

se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
ALLAMAI
QBALOPENUNI
VERSI
TY
Assi
gnmentNo.2
Name:
Umme-
Habi
ba
Rol
lno.:
Cb646471
Regi
str
ati
onNo.:
20pst
03979
Fat
herName:
MaqsoodAhmad
Lev
el:
B.Ed(
1.5y
ear
s)
Cour
seCode:
8614-
2
Cour
seTi
tl
e:Educat
ional
Stat
ist
ics

ASSI
GNMENTNo.2
Q.
1 Whati
sthe i
mpor
tance ofi
nfer
ent
ialst
ati
sti
cs i
n educat
ionalr
esear
ch? El
abor
ate wi
th
exampl
es.
Answer
:
I
neducat
ionalr
esear
ch,i
tisnev
erpossi
blet
osampl
eal
loft
hepeopl
ethatwewantt
odr
aw a
concl
usi
onabout
.Forex
ampl
e,t
othepur
poseofar
esear
chst
udymaybet
odet
ermi
nei
fanewway
oft
eachi
ngmat
hemat
icsi
mpr
ovesmat
hemat
icalachi
evementf
oral
lchi
l
dreni
nPr
imar
y1.Howev
er,
i
twoul
dbei
mpossi
blet
otestal
lchi
l
dreni
nPr
imar
y1becauseoft
ime,
resour
ces,
andot
herl
ogi
sti
cal
f
act
ors.I
nst
ead,
ther
esear
cherchoosesasampl
eoft
hepopul
ati
ont
oconductast
udy
.Howev
er,
the
r
esear
cherdoesnotwantt
oli
mitherconcl
usi
onst
oonl
ythef
ew chi
l
drenwhopar
ti
cipat
edi
nthe
st
udy
.Inst
ead,shewant
stosayt
hatbecausehernew wayoft
eachi
ngmat
hemat
icsi
mpr
oved
mat
hemat
icalachi
evementf
ort
he chi
l
dren i
n hersampl
e,i
twi
l
lal
so i
mpr
ove mat
hemat
ical
achi
evementf
oral
lchi
l
dren i
n Pr
imar
y1.Thus,t
he r
esear
cherwant
sto dr
aw concl
usi
ons -
ori
nfer
ences-aboutt
heent
ir
epopul
ati
onbasedont
her
esul
tsf
rom t
hesampl
e.
Thepur
poseofi
nfer
ent
ialst
ati
sti
csi
stodet
ermi
newhet
hert
hef
indi
ngsf
rom t
hesampl
ecan
gener
ali
ze-orbeappl
i
ed-t
otheent
ir
epopul
ati
on.Ther
ewi
l
lal
way
sbedi
ff
erencesi
nscor
es
bet
weengr
oupsi
nar
esear
chst
udy
.Forexampl
e,i
magi
net
hef
oll
owi
ngscor
esonWAEC f
rom
st
udent
swhoat
tendedei
theragov
ernmentorapr
ivat
eschool
.(Not
ethatt
hedat
awascr
eat
edbya
r
andom numbergener
atorandt
husdoesnotr
efl
ectact
ual
scor
es.
)

Cour
se Gov
ernmentPr
ivat
e

Engl
i
sh 24.
13 28.
18

1
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
Mat
hemat
ics 58.
90 27.
87

I
ntegr
atedSci
ence 61.
19 47.
89

Chemi
str
y 34.
38 34.
25

Soci
alSt
udi
es 99.
10 86.
97

CRK/
IRK 70.
90 70.
53

I
nal
lcour
sesl
i
stedi
nTabl
e,t
her
eisadi
ff
erencei
ntheav
eragescor
esbet
weent
hegov
ernmentand
pr
ivat
eschool
s.Howev
er,
whi
chdi
ff
erencesar
elar
geenought
obesi
gni
fi
cant
?Inmat
hemat
ics,
ther
e
i
sacl
eardi
ff
erencei
nthemeanscor
es.Ont
heot
herhand,t
hedi
ff
erencebet
weengov
ernmentand
pr
ivat
eschool
sonCRK/
IRK i
slesst
han1poi
nt.Thi
scl
ear
lydoesnotr
epr
esentameani
ngf
ul
di
ff
erencebet
weent
het
wot
ypesofschool
s.How bi
gmustt
hedi
ff
erencebei
nor
dert
oconcl
ude
t
hatt
her
eisameani
ngf
uldi
ff
erencei
ntheacademi
cper
for
mancebet
weengov
ernmentandpr
ivat
e
school
s?
I
nfer
ent
ial
stat
ist
icsmustbeusedt
odet
ermi
newhet
hert
hedi
ff
erencebet
weent
het
wogr
oupsi
nthe
sampl
eisl
argeenought
obeabl
etosayt
hatt
hef
indi
ngsar
esi
gni
fi
cant
.Ift
hef
indi
ngsar
eindeed
si
gni
fi
cant
,thent
heconcl
usi
onscanbeappl
i
ed,
orgener
ali
zed,
tot
heent
ir
epopul
ati
on.Ont
heot
her
hand,i
fthedi
ff
erencebet
weent
hegr
oupsi
sver
ysmal
l
,thent
hef
indi
ngsar
enotsi
gni
fi
cantand
t
her
efor
ewer
esi
mpl
yther
esul
tofchance.
Ret
urni
ng t
othe WAEC i
l
lust
rat
ion,t
hisr
esear
ch f
ound t
hatt
her
eis a di
ff
erence i
n CRK/
IRK
per
for
mancei
noursampl
e:gov
ernmentschool
shadasl
i
ght
lyhi
ghermeanscor
ethanpr
ivat
e
school
s.Howev
er,af
terconduct
ing i
nfer
ent
ialst
ati
sti
cs,wewi
l
lfi
nd t
hatt
hedi
ff
erencei
snot
si
gni
fi
cant
.Thi
smeanst
hatt
hedi
ff
erencebet
weengov
ernmentandpr
ivat
eschool
sinoursampl
e
wast
oosmal
ltoconcl
udet
hatanact
ualdi
ff
erenceexi
stsi
ntheent
ir
epopul
ati
onofgov
ernmentand
pr
ivat
eschool
s.Thi
sti
nydi
ff
erencei
nCRK/
IRK per
for
mancewassi
mpl
yduet
ochance.I
fwe
col
l
ect
eddat
afr
om anot
hersampl
eofgov
ernmentandpr
ivat
eschool
s,wewi
l
lli
kel
yfi
ndt
he
opposi
teoft
hissmal
ldi
ff
erence.
Ont
heot
herhand,af
terconduct
ingi
nfer
ent
ialst
ati
sti
cs,i
magi
net
hatwef
indt
hatt
hedi
ff
erencei
n
mat
hemat
icsper
for
mance i
ssi
gni
fi
cant
.Ther
efor
e,t
he r
esul
tsf
rom oursampl
e -gov
ernment
school
shav
ehi
gherper
for
mancei
nmat
hemat
icst
hanpr
ivat
eschool
s-canal
sobeappl
i
edt
othe
ent
ir
epopul
ati
on.Thus,
weconcl
udet
hati
ngener
al,
gov
ernmentschool
sout
per
for
m pr
ivat
eschool
s
i
nmat
hemat
ics.Thi
smeanst
hati
fwecol
l
ectdat
afr
om anot
hersampl
e,wear
eal
mostguar
ant
eed
t
ofi
ndt
heexactsamer
esul
t.
Whencal
cul
ati
ngi
nfer
ent
ialst
ati
sti
cs,
thekeyst
ati
sti
cist
hepst
ati
sti
c.Thi
sp-
val
uei
sthepr
obabi
l
ity

2
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
t
hatt
her
esul
tisduet
ochance.Thep-
val
uecanr
angef
rom 0.
000t
o1.
000.Thel
argerpi
s,t
hemor
e
l
i
kel
yther
esul
tsar
eduet
ochance.I
fpi
s0.
500,t
hepr
obabi
l
ityt
hatt
her
esul
tisduet
ochancei
s5
outof10.Thi
smeanst
hatt
her
eisanequalpr
obabi
l
ityt
hatt
her
esul
tissi
gni
fi
canti
nthepopul
ati
on
ort
hatt
her
esul
tisduet
ochance.Apof0.
850meanst
hatt
hepr
obabi
l
ityt
hatt
her
esul
tisduet
o
chancei
s85outof100,meani
ngt
hatt
hef
indi
ngi
smostl
i
kel
yduet
ochance.Apof0.
050means
t
hepr
obabi
l
ityt
hatt
her
esul
tisduet
ochancei
s5outof100.
Resear
cher
swantt
obef
air
lycer
tai
noft
hei
rconcl
usi
ons,sot
heyhav
edeci
dedt
hatt
hepr
obabi
l
ity
t
hatt
her
esul
tisduet
ochanceshoul
dbev
erysmal
l-l
esst
han5outof100.Thi
siswhyt
hest
andar
d
i
nresear
chi
sthatt
hepmustbei
slesst
han.
05f
ort
her
esul
tst
obesi
gni
fi
cant
.(Thi
scut
-of
fpoi
ntf
or
pi
sof
tent
imescal
l
edt
heal
pha,
somet
imesst
atedt
hatal
phai
slesst
han.
05.
)
Recal
lfr
om Wr
it
ing Resear
chHy
pot
hesest
hatr
esear
chhy
pot
hesesar
egener
all
yst
ated asnul
l
hy
pot
heses:
Ther
eis no si
gni
fi
cantdi
ff
erence bet
ween gov
ernmentand pr
ivat
e school
s on
mat
hemat
icsper
for
mance.I
fthecal
cul
atedp-
val
uei
slesst
han0.
050,t
hent
henul
lhy
pot
hesi
s
i
srej
ect
ed.I
fthest
atement"
Ther
eisnosi
gni
fi
cantdi
ff
erence"i
srej
ect
ed,t
hismeanst
hatt
her
eisa
si
gni
fi
cantdi
ff
erence.I
tcant
her
efor
ebeconcl
udedt
hatadi
ff
erenceexi
stsbet
weent
het
wogr
oups
i
nthepopul
ati
on,notj
ustt
hesampl
e.Ont
heot
herhand,i
fpi
sgr
eat
ert
han0.
050,t
hismeanst
hat
t
he nul
lhy
pot
hesi
sisr
etai
ned.Thi
s means t
he nul
lhy
pot
hesi
s,as st
ated,i
s accur
ate:i
nthe
popul
ati
on,
ther
eisnosi
gni
fi
cantdi
ff
erence.
Att
hispoi
nt,
thedat
ashoul
dbecodedandt
otalscor
escal
cul
atedf
oreachv
ari
abl
e.I
tisnowt
imet
o
cal
cul
atet
he i
nfer
ent
ialst
ati
sti
cs.Ihi
ghl
yrecommend usi
ngVassar
Stat
s,a f
ree websi
tet
hat
cal
cul
atesst
ati
sti
csf
oral
lbutt
hemostcompl
exst
udi
es.Thespeci
fi
cst
ati
sti
ctocal
cul
atef
oreach
r
esear
chhy
pot
hesi
sshoul
dhav
ebeenal
readyi
dent
if
iedi
nMet
hodofDat
aAnal
ysi
sst
ep.I
fthi
shas
beendone,t
hedat
aanal
ysi
sshoul
dtakel
esst
hananhourt
ocompl
ete.Eacht
ypeofi
nfer
ent
ial
st
ati
sti
cissl
i
ght
lydi
ff
erent
,sor
eadbel
owf
orhowt
ocal
cul
ateeachst
ati
sti
c.
t
-t
est
Thet
-t
estshoul
dbeusedwhencompar
ingt
wogr
oupsonadependentv
ari
abl
e.Thus,
at-
testwi
l
lbe
usedt
ocompar
eat
reat
mentgr
oupt
oacont
rolgr
ouport
ocompar
emal
esandf
emal
es.Forsakeof
i
l
lust
rat
ion,r
eferbackt
othecompar
isonofgov
ernmentandpr
ivat
eschool
sonWAECscor
es.To
cal
cul
atet
het
-t
est
,thedat
amustbesor
tedaccor
dingt
othei
ndependentv
ari
abl
e,i
nthi
scaset
ype
ofschool
.Inot
herwor
ds,al
loft
hegov
ernmentschoolscor
esar
egr
oupedt
oget
her
,andal
loft
he
pr
ivat
eschoolscor
esar
egr
oupedt
oget
her
.InVassar
Stat
s,cl
i
ckont
het
-t
est
,thenTwo-
sampl
et-
test
f
ori
ndependentorcor
rel
atedsampl
es.I
nthenew scr
een,cl
i
ckI
ndependentSampl
es.(
Cor
rel
ated
sampl
eswoul
dbef
orst
udi
eswher
ethet
wogr
oupswer
emat
chedonanot
herr
elev
antv
ari
abl
e,such
asi
ntel
l
igence,
orwhenbot
hscor
escamef
rom t
hesamegr
oupofpeopl
e,suchascompar
ingEngl
i
sh

3
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
andmat
hsscor
eswi
thi
nthegov
ernmentschool
s.)Thenent
ert
hemat
hemat
icsscor
esf
oreach
st
udenti
nthegov
ernmentschoolgr
oupi
nonesampl
ear
ea(
Sampl
eA)andt
hemat
hemat
icsscor
es
f
oreachst
udenti
nthepr
ivat
eschool
groupi
ntheot
her
.Cl
i
ckCal
cul
ateandt
her
esul
tsshoul
dpopup.
Scr
olldownt
hescr
een.Thef
ir
stst
ati
sti
ctol
ookati
sthep:t
wo-
tai
l
ed.I
fthi
spi
sgr
eat
ert
han0.
050,
t
hent
henul
lhy
pot
hesi
sisr
etai
ned;
ther
esul
tisnotsi
gni
fi
cant
.Ift
her
esul
tisnotsi
gni
fi
cant
,anal
ysi
s
andi
nter
pret
ati
oni
sfi
nishedbecauset
her
eisnosi
gni
fi
cantdi
ff
erencebet
weengr
oups.
I
fthi
spi
slesst
han0.
050,t
hent
henul
lhy
pot
hesi
sisr
eject
edandt
her
esul
tissi
gni
fi
cant
.Ift
hisi
s
si
gni
fi
cant
,thent
henextst
epi
stol
ookatt
hemeanscor
eforeachgr
oup(
scr
ollbackupt
hepagef
or
t
his)
.Basedont
hosemeanscor
es,whi
chgr
ouphadt
hehi
ghermat
hemat
icsscor
es:gov
ernmentor
pr
ivat
eschool
s?Thi
slastst
epofexami
ningt
hemeanscor
esi
sof
tenov
erl
ookedi
ndat
aanal
ysi
s,but
i
tisv
eryi
mpor
tantt
oident
if
ywhi
chgr
ouphadt
hehi
ghestmeanscor
e.I
tisi
mpor
tantt
orecor
dthe
meansandst
andar
ddev
iat
ionsf
ort
het
wogr
oups,t
het
,df
,andt
wo-
tai
l
edp.Cl
i
ckt
heResetbut
ton
andmov
etot
henextr
esear
chhy
pot
hesi
s.
Anal
ysi
sofVar
iance(
ANOVA)
Whencompar
ingt
hreeormor
egr
oupsononedependentv
ari
abl
e,anAnal
ysi
sofVar
iancei
sthe
st
ati
sti
cst
ouse.Ther
ear
etwobasi
cty
pesofANOVAst
hatcanbeused.
One-
wayANOVA:
Aone-
wayANOVAcompar
esmul
ti
plegr
oupsont
hesamev
ari
abl
e.Forexampl
e,
per
hapst
her
esear
cherdeci
dest
odi
vi
depr
ivat
eschool
sint
orel
i
giouspr
ivat
eandsecul
arpr
ivat
e.
Now,t
her
ear
ethr
eegr
oupst
obecompar
ed:gov
ernmentschool
s,r
eli
giouspr
ivat
e,andsecul
ar
pr
ivat
e.Aone-
wayANOVAi
snow necessar
y.Tocal
cul
atet
heone-
wayANOVA,t
hedat
amustbe
sor
tedaccor
dingt
othei
ndependentv
ari
abl
e-agai
n,schoolt
ype.I
nVassar
Stat
s,cl
i
ckonANOVA,
t
henOne-
WayANOVA.Thenent
ert
henumberofsampl
es(
akat
henumberofgr
oups;
int
hisexampl
e,
3)
.Then cl
i
ck I
ndependentSampl
es.Ent
ert
he mat
hemat
ics scor
es f
oreach st
udenti
nthe
appr
opr
iat
ecol
umn.Forexampl
e,ent
ergov
ernmentst
udent
s'scor
esi
nSampl
eA,r
eli
giouspr
ivat
e
st
udent
s'scor
esi
nSampl
eB,
andsecul
arpr
ivat
est
udent
s'scor
esi
nSampl
eC.Thencl
i
ckCal
cul
ate.
Scr
oll
downt
hescr
een.Agai
n,t
hef
ir
stst
ati
sti
ctol
ookati
sthepi
ntheANOVAsummar
ytabl
e.Agai
n,
i
fthi
spi
sgr
eat
ert
han0.
050,t
hent
henul
lhy
pot
hesi
sisr
etai
ned;t
her
esul
tisnotsi
gni
fi
cant
.Ift
he
r
esul
tisnotsi
gni
fi
cant
,anal
ysi
sand i
nter
pret
ati
on i
sfi
nished becauset
her
eisno si
gni
fi
cant
di
ff
erencebet
weengr
oups.
I
fthi
spi
slesst
han0.
050,t
hent
her
esul
tissi
gni
fi
cant
.Thi
sonl
ymeans,howev
er,t
hatt
her
eisa
si
gni
fi
cantdi
ff
erencebet
weengr
oupssomewher
e,nott
hatt
her
eisasi
gni
fi
cantdi
ff
erencebet
ween
al
lgr
oups.I
tispossi
blet
hatgov
ernmentst
udent
swer
esi
gni
fi
cant
lyhi
ghert
hanr
eli
giouspr
ivat
eand
secul
arpr
ivat
est
udent
s,butt
her
ear
enosi
gni
fi
cantdi
ff
erencesbet
weenr
eli
giouspr
ivat
eandsecul
ar
pr
ivat
est
udent
s.Downatt
hebot
tom oft
hescr
eeni
sther
esul
tofTukey
'sHSD(
Honest
lySi
gni
fi
cant

4
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
Di
ff
erence)t
est
.Thi
stesti
dent
if
ieswhi
chdi
ff
erencesar
ereal
l
ysi
gni
fi
cant
.Iti
simpor
tantt
orecor
d
t
hemeansandst
andar
ddev
iat
ionsf
oral
lgr
oups,t
heANOVAsummar
ytabl
e,andt
her
esul
tsof
Tukey
'sHSD.Cl
i
ckt
heResetbut
tonandmov
etot
henextr
esear
chhy
pot
hesi
s.
Fact
ori
alANOVA:
Thef
act
ori
alANOVA compar
est
heef
fectofmul
ti
plei
ndependentv
ari
abl
esononedependent
v
ari
abl
e.Forexampl
e,a2x3f
act
ori
alANOVAcoul
dcompar
etheef
fect
sofgenderandschool
typeon
academi
cper
for
mance.Thef
ir
sti
ndependentv
ari
abl
e,gender
,hast
wol
evel
s(mal
eandf
emal
e)and
t
hesecondi
ndependentv
ari
abl
e,schoolt
ype,hast
hreel
evel
s(gov
ernment
,rel
i
giouspr
ivat
e,and
secul
arpr
ivat
e),hence2x3(
read"
twobyt
hree"
).Fact
ori
alANOVAscanal
sobecal
cul
atedon
Vassar
Stat
s(cl
i
ckonANOVAt
henonTwo-
wayf
act
ori
alANOVAf
ori
ndependentsampl
es)
.Howev
er,
t
hisi
nter
pret
ati
oni
sabi
tmor
ecompl
exsopl
easeseeanexper
tst
ati
sti
ciant
ohel
pwi
thi
nter
pret
ing
t
her
esul
ts.
Anal
ysi
sofCov
ari
ance(
ANCOVA)
:
Whenusi
ngapr
e-postt
estr
esear
chdesi
gn,
theAnal
ysi
sofCov
ari
anceal
l
owsacompar
isonofpost
-
t
estscor
eswi
thpr
e-t
estscor
esf
act
oredout
.Forexampl
e,i
fcompar
ingat
reat
mentandcont
rol
gr
oup on achi
evementmot
ivat
ion wi
th a pr
e-postt
estdesi
gn,t
he ANCOVA wi
l
lcompar
ethe
t
reat
mentandcont
rol
groups'
post
-t
estscor
esbyst
ati
sti
cal
l
yadj
ust
ingf
ort
hepr
e-t
estscor
es.Foran
ANCOVA,y
oumusthav
epr
e-andpost
-t
estscor
esf
orev
eryper
soni
nthesampl
e,andt
hesescor
es
mustbesor
tedbyt
hegr
oup(
akat
reat
mentandcont
rol
group)
.
Tocal
cul
ateanANCOVAwi
thVassar
Stat
s,cl
i
ckonANCOVA.ThenVassar
Stat
swi
l
laskf
ort
hek.The
ki
sthenumberofgr
oups.I
fther
eisonl
yonet
reat
mentandonecont
rolgr
oup,
thenk=2.Cl
i
ckont
he
cor
rectkf
ordat
aimpor
t.Ther
ear
etwot
hingst
obeari
nmi
ndwhendoi
ngANCOVAwi
thVassar
Stat
.
I
twi
l
laskf
ort
heconcomi
tantv
ari
abl
eandt
hedependentv
ari
abl
e.Theconcomi
tantv
ari
abl
e(CV)i
s
t
hev
ari
abl
ethatshoul
dbecont
rol
l
edf
or.I
nthecaseofapr
e-postt
estdesi
gn,t
heconcomi
tant
v
ari
abl
eist
hepr
e-t
est
.Thedependentv
ari
abl
e(DV)i
sthev
ari
abl
ethaty
out
hinkhasbeenaf
fect
ed
byt
hei
ndependentv
ari
abl
e.I
nthecaseofapr
e-postt
estdesi
gn,
thedependentv
ari
abl
eist
hepost
-
t
est
.TouseVassar
Stat
s,i
tisi
mpor
tantt
hatt
heCVandt
heDVar
esi
de-
by-
sidef
oreachoft
het
wo
gr
oups.Thenent
ert
heCVandDVi
ntot
hecor
rectcol
umnsandcl
i
ckCal
cul
ate.
Scr
olldownt
hescr
een.Justasbef
ore,t
hef
ir
stst
ati
sti
ctol
ookati
sthepi
ntheANCOVAsummar
y
t
abl
e.I
fthi
spi
slesst
han0.
050,t
hent
henul
lhy
pot
hesi
sisr
eject
edandt
her
esul
tissi
gni
fi
cant
.
Ther
ear
etwoset
sofmeanst
hatar
eimpor
tantt
ounder
standi
nanANCOVA.Fi
rst
,theObser
ved
Meansar
etheact
ualmeansf
ort
hedependentv
ari
abl
e(post
-t
est
).Thent
heAdj
ust
edMeansar
ethe
meanst
hathav
ebeenst
ati
sti
cal
l
ymani
pul
atedbasedont
hepr
e-t
estscor
es.A si
mpl
ewayt
o
i
magi
net
hisi
sthatt
heANCOVAst
ati
sti
cal
l
yfor
cest
hepr
e-t
estscor
est
obeequalbet
weent
het
wo

5
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
gr
oups(
meani
ngt
hatt
het
wogr
oupsar
enowequalatt
hest
artoft
hest
udy
),andt
henr
e-cal
cul
ates
t
hepost
-t
estscor
esbasedont
headj
ust
edpr
e-t
estscor
es.I
tisi
mpor
tantt
orecor
dtheobser
ved
means,adj
ust
edmeans,andst
andar
ddev
iat
ionsf
oral
lgr
oupsandt
heACNOVAsummar
ytabl
e.
Whencr
eat
ingt
het
abl
esi
nthenextst
ep,r
epor
tbot
htheObser
vedandAdj
ust
edMeans.Howev
er,
makeanyf
igur
esbasedwi
tht
heAdj
ust
edMeans.Addanot
etot
hef
igur
esot
hatr
eader
sar
ecl
ear
t
hatt
hesear
eAdj
ust
edMeans.Cl
i
ckt
heResetbut
tonandmov
etot
henextr
esear
chhy
pot
hesi
s.
Cor
rel
ati
on:
Cor
rel
ati
onsshoul
dbecal
cul
atedt
oexami
net
her
elat
ionshi
pbet
weent
wov
ari
abl
eswi
thi
nthesame
gr
oupofpar
ti
ci
pant
s.Forexampl
e,t
hecor
rel
ati
onwoul
dquant
if
yther
elat
ionshi
pbet
weenacademi
c
achi
evementandachi
evementmot
ivat
ion.Tocal
cul
ateacor
rel
ati
on,y
oumusthav
escor
esf
ort
wo
v
ari
abl
esf
orev
erypar
ti
cipanti
nthesampl
e.Tocal
cul
ateacor
rel
ati
oni
nVassar
Stat
s,cl
i
ckon
Cor
rel
ati
on&Regr
essi
on,t
henBasi
cLi
nearCor
rel
ati
onandRegr
essi
on,Dat
a-I
mpor
tVer
sion.Ent
er
t
het
otal
scor
esf
ort
het
wov
ari
abl
esandcl
i
ckCal
cul
ate.
Scr
olldownt
hescr
een.Agai
n,t
hef
ir
stst
ati
sti
ctol
ookati
sthep:t
wo-
tai
l
ed.Thenul
lhy
pot
hesesf
or
cor
rel
ati
onsst
ate,
Ther
eisnosi
gni
fi
cantr
elat
ionshi
pbet
weenmat
hemat
icsandEngl
i
shachi
evement
.
I
fthepi
sgr
eat
ert
han0.
050,t
hent
henul
lhy
pot
hesi
sisr
etai
ned;t
her
eisnosi
gni
fi
cantr
elat
ionshi
p
bet
weenv
ari
abl
es.I
fther
esul
tisnotsi
gni
fi
cant
,anal
ysi
sandi
nter
pret
ati
oni
sfi
nishedbecauset
her
e
i
snosi
gni
fi
cantr
elat
ionshi
p.
I
fthi
spi
slesst
han0.
050,
thent
henul
lhy
pot
hesi
sisr
eject
edandt
hecor
rel
ati
oni
ssi
gni
fi
cant
.Ift
he
cor
rel
ati
oni
ssi
gni
fi
cant
,thent
henextst
epi
stol
ookatt
hecor
rel
ati
oni
tsel
f,sy
mbol
i
zedbyr
.For
mor
einf
ormat
iononhowt
oint
erpr
ett
hecor
rel
ati
on,
cli
ckonMet
hodofDat
aAnal
ysi
s.I
tisi
mpor
tant
t
orecor
dthemeansandst
andar
ddev
iat
ionsf
ort
het
wov
ari
abl
es,t
het
,df
,two-
tai
l
edp,andr
.Cl
i
ck
t
he"
Reset
"but
tonandmov
etot
henextr
esear
chhy
pot
hesi
s.
Q.
2 Whatdomeanbyhy
pot
hesi
stest
ing?Br
ief
lyst
atet
hel
ogi
cbehi
ndhy
pot
hesi
stest
ing?
Answer
:
Hy
pot
hesi
stest
ing i
san acti
n st
ati
sti
cswher
ebyan anal
ystt
est
san assumpt
ion r
egar
ding a
popul
ati
onpar
amet
er.Themet
hodol
ogyempl
oyedbyt
heanal
ystdependsont
henat
ureoft
hedat
a
usedandt
her
easonf
ort
heanal
ysi
s.
Hy
pot
hesi
stest
ingi
susedt
oassesst
hepl
ausi
bil
i
tyofahy
pot
hesi
sbyusi
ngsampl
edat
a.Suchdat
a
maycomef
rom al
argerpopul
ati
on,
orf
rom adat
a-gener
ati
ngpr
ocess.Thewor
d"popul
ati
on"wi
l
lbe
usedf
orbot
hoft
hesecasesi
nthef
oll
owi
ngdescr
ipt
ions.
 Hy
pot
hesi
stest
ingi
susedt
oassesst
hepl
ausi
bil
i
tyofahy
pot
hesi
sbyusi
ngsampl
edat
a.
 Thet
estpr
ovi
desev
idenceconcer
ningt
hepl
ausi
bil
i
tyoft
hehy
pot
hesi
s,gi
vent
hedat
a.
 St
ati
sti
calanal
yst
stestahy
pot
hesi
sbymeasur
ingandexami
ningar
andom sampl
eoft
he

6
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
popul
ati
onbei
nganal
yzed.
I
nhy
pot
hesi
stest
ing,
ananal
ystt
est
sast
ati
sti
calsampl
e,wi
tht
hegoalofpr
ovi
dingev
idenceont
he
pl
ausi
bil
i
tyoft
henul
lhy
pot
hesi
s.
St
ati
sti
calanal
yst
stesta hy
pot
hesi
s bymeasur
ing and exami
ning a r
andom sampl
e oft
he
popul
ati
onbei
nganal
yzed.Al
lanal
yst
susear
andom popul
ati
onsampl
etot
estt
wo di
ff
erent
hy
pot
heses:
thenul
lhy
pot
hesi
sandt
heal
ter
nat
ivehy
pot
hesi
s.
Thenul
lhy
pot
hesi
sisusual
l
yahy
pot
hesi
sofequal
i
tybet
weenpopul
ati
onpar
amet
ers;e.
g.,anul
l
hy
pot
hesi
smayst
atet
hatt
hepopul
ati
onmeanr
etur
nisequalt
ozer
o.Theal
ter
nat
ivehy
pot
hesi
sis
ef
fect
ivel
ytheopposi
teofanul
lhy
pot
hesi
s(e.
g.,t
hepopul
ati
onmeanr
etur
nisnotequalt
ozer
o).
Thus,
theyar
emut
ual
l
yexcl
usi
ve,
andonl
yonecanbet
rue.Howev
er,
oneoft
het
wohy
pot
heseswi
l
l
al
way
sbet
rue.
4St
epsofHy
pot
hesi
sTest
ing:
Al
lhy
pot
hesesar
etest
edusi
ngaf
our
-st
eppr
ocess:
1.Thef
ir
stst
epi
sfort
heanal
ystt
ost
atet
het
wohy
pot
hesessot
hatonl
yonecanber
ight
.
2.Thenextst
epi
stof
ormul
ateananal
ysi
spl
an,
whi
chout
li
neshowt
hedat
awi
l
lbeev
aluat
ed.
3.Thet
hir
dst
epi
stocar
ryoutt
hepl
anandphy
sical
l
yanal
yzet
hesampl
edat
a.
4.Thef
our
thandf
inalst
epi
stoanal
yzet
her
esul
tsandei
therr
ejectt
henul
lhy
pot
hesi
s,orst
ate
t
hatt
henul
lhy
pot
hesi
sispl
ausi
ble,
giv
ent
hedat
a.
I
f,f
orexampl
e,aper
sonwant
stot
estt
hatapennyhasexact
lya50%chanceofl
andi
ngonheads,
t
henul
lhy
pot
hesi
swoul
dbet
hat50%i
scor
rect
,andt
heal
ter
nat
ivehy
pot
hesi
swoul
dbet
hat50%i
s
notcor
rect
.
Mat
hemat
ical
l
y,t
henul
lhy
pot
hesi
swoul
dber
epr
esent
edasHo:P=0.
5.Theal
ter
nat
ivehy
pot
hesi
s
woul
dbedenot
edas"
Ha"andbei
dent
icalt
othenul
lhy
pot
hesi
s,exceptwi
tht
heequalsi
gnst
ruck-
t
hrough,
meani
ngt
hati
tdoesnotequal
50%.
Ar
andom sampl
eof100coi
nfl
i
psi
staken,andt
henul
lhy
pot
hesi
sist
hent
est
ed.I
fiti
sfoundt
hat
t
he100coi
nfl
i
pswer
edi
str
ibut
edas40headsand60t
ail
s,t
heanal
ystwoul
dassumet
hatapenny
doesnothav
ea50%chanceofl
andi
ngonheadsandwoul
drej
ectt
henul
lhy
pot
hesi
sandacceptt
he
al
ter
nat
ivehy
pot
hesi
s.
I
f,ont
heot
herhand,
ther
ewer
e48headsand52t
ail
s,t
heni
tispl
ausi
blet
hatt
hecoi
ncoul
dbef
air
andst
il
lpr
oducesuchar
esul
t.I
ncasessuchast
hiswher
ethenul
lhy
pot
hesi
sis"
accept
ed,
"the
anal
ystst
atest
hatt
hedi
ff
erencebet
weent
heexpect
edr
esul
ts(
50headsand50t
ail
s)andt
he
obser
vedr
esul
ts(
48headsand52t
ail
s)i
s"expl
ainabl
ebychanceal
one.
"
Q.
3 Br
ief
lyexpl
ain whaty
ou under
stand by“
cor
rel
ati
on”
?Wr
it
e down wher
e and whyt
o use
cor
rel
ati
on?Wr
it
e down t
he consi
der
ati
on t
hatshoul
d be kepti
n mi
nd whi
l
eint
erpr
eti
ng

7
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
cor
rel
ati
on.
Answer
:
Regr
essi
onanal
ysi
sinv
olv
esi
dent
if
yingt
her
elat
ionshi
pbet
weenadependentv
ari
abl
eandoneor
mor
eindependentv
ari
abl
es.A modeloft
her
elat
ionshi
pishy
pot
hesi
zed,andest
imat
esoft
he
par
amet
erv
aluesar
eusedt
odev
elopanest
imat
edr
egr
essi
onequat
ion.Var
ioust
est
sar
ethen
empl
oyed t
o det
ermi
ne i
fthe modeli
s sat
isf
act
ory
.Ift
he modeli
s deemed sat
isf
act
ory
,the
est
imat
edr
egr
essi
onequat
ioncanbeusedt
opr
edi
ctt
hev
alueoft
hedependentv
ari
abl
egi
ven
v
aluesf
ort
hei
ndependentv
ari
abl
es.
Regr
essi
onmodel
.
I
nsi
mpl
eli
nearr
egr
essi
on,t
hemodelusedt
odescr
ibet
her
elat
ionshi
pbet
weenasi
ngl
edependent
v
ari
abl
eyandasi
ngl
eindependentv
ari
abl
exi
sy=a0+a1x+k.a0anda1ar
eref
err
edt
oast
hemodel
par
amet
ers,andi
sapr
obabi
l
ist
icer
rort
ermt
hataccount
sfort
hev
ari
abi
l
ityi
nyt
hatcannotbe
expl
ainedbyt
hel
i
nearr
elat
ionshi
pwi
thx.I
ftheer
rort
erm wer
enotpr
esent
,themodelwoul
dbe
det
ermi
nist
ic;
int
hatcase,
knowl
edgeoft
hev
alueofxwoul
dbesuf
fi
cientt
odet
ermi
net
hev
alueofy
.

Leastsquar
esmet
hod:
Ei
therasi
mpl
eormul
ti
pler
egr
essi
onmodeli
sini
ti
all
yposed asahy
pot
hesi
sconcer
ning t
he
r
elat
ionshi
pamongt
hedependentandi
ndependentv
ari
abl
es.Thel
eastsquar
esmet
hodi
sthemost
wi
del
yusedpr
ocedur
efordev
elopi
ngest
imat
esoft
hemodel
par
amet
ers.
Asani
l
lust
rat
ionofr
egr
essi
onanal
ysi
sandt
hel
eastsquar
esmet
hod,supposeauni
ver
sit
ymedi
cal
cent
rei
sinv
est
igat
ingt
her
elat
ionshi
pbet
weenst
ressandbl
oodpr
essur
e.Assumet
hatbot
hast
ress
t
estscor
eandabl
oodpr
essur
ereadi
nghav
ebeenr
ecor
dedf
orasampl
eof20pat
ient
s.Thedat
aar
e
showngr
aphi
cal
l
yint
hef
igur
ebel
ow,cal
l
edascat
terdi
agr
am.Val
uesoft
hei
ndependentv
ari
abl
e,
st
resst
estscor
e,ar
egi
venont
hehor
izont
alaxi
s,andv
aluesoft
hedependentv
ari
abl
e,bl
ood
pr
essur
e,ar
eshownont
hev
ert
icalaxi
s.Thel
i
nepassi
ngt
hrought
hedat
apoi
ntsi
sthegr
aphoft
he
est
imat
edr
egr
essi
onequat
ion:y=42.
3+0.
49x
.Thepar
amet
erest
imat
es,b0=42.
3andb1=0.
49,
wer
eobt
ainedusi
ngt
hel
eastsquar
esmet
hod.

8
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021

Cor
rel
ati
on.
Cor
rel
ati
onandr
egr
essi
onanal
ysi
sar
erel
atedi
nthesenset
hatbot
hdealwi
thr
elat
ionshi
psamong
v
ari
abl
es.Thecor
rel
ati
oncoef
fi
cienti
sameasur
eofl
i
nearassoci
ati
onbet
weent
wov
ari
abl
es.
Val
uesoft
hecor
rel
ati
oncoef
fi
cientar
eal
way
sbet
ween-
1and+1.Acor
rel
ati
oncoef
fi
cientof+1
i
ndi
cat
est
hatt
wov
ari
abl
esar
eper
fect
lyr
elat
edi
naposi
ti
vel
i
nearsense,
acor
rel
ati
oncoef
fi
cientof
-
1indi
cat
est
hatt
wov
ari
abl
esar
eper
fect
lyr
elat
edi
nanegat
ivel
i
nearsense,andacor
rel
ati
on
coef
fi
cientof0i
ndi
cat
est
hatt
her
eisnol
i
nearr
elat
ionshi
pbet
weent
het
wov
ari
abl
es.Forsi
mpl
e
l
i
nearr
egr
essi
on,t
he sampl
e cor
rel
ati
on coef
fi
cienti
sthe squar
erootoft
he coef
fi
cientof
det
ermi
nat
ion,wi
tht
hesi
gnoft
hecor
rel
ati
oncoef
fi
cientbei
ngt
hesameast
hesi
gnofb1,t
he
coef
fi
cientofx1i
ntheest
imat
edr
egr
essi
onequat
ion.
Nei
therr
egr
essi
onnorcor
rel
ati
onanal
ysescanbei
nter
pret
ed asest
abl
i
shi
ng cause-
and-
eff
ect
r
elat
ionshi
ps.Theycani
ndi
cat
eonl
yhowort
owhatext
entv
ari
abl
esar
eassoci
atedwi
theachot
her
.
Thecor
rel
ati
oncoef
fi
ci
entmeasur
esonl
ythedegr
eeofl
i
nearassoci
ati
onbet
weent
wov
ari
abl
es.Any
concl
usi
onsaboutacause-
and-
eff
ectr
elat
ionshi
pmustbebasedont
hej
udgmentoft
heanal
yst
.
Q.
4 Whenwi
l
lyouuseANOVAi
nyourr
esear
ch?Wr
it
edownt
hel
ogi
cbehi
ndusi
ngANOVA.
Answer
:
I
nsomedeci
sion-
maki
ngsi
tuat
ions,t
hesampl
edat
amaybedi
vi
dedi
ntov
ari
ousgr
oupsi
.e.t
he
sampl
emaybesupposedt
ohav
econsi
stedofk-
subsampl
es.Ther
ear
eint
erestl
i
esi
nexami
ning
whet
hert
het
otalsampl
ecanbeconsi
der
edashomogenousort
her
eissomei
ndi
cat
iont
hatsub-
sampl
eshav
ebeendr
awnf
rom di
ff
erentpopul
ati
ons.So,
int
hesesi
tuat
ions,
wehav
etocompar
ethe
meanv
aluesofv
ari
ousgr
oups,
wit
hrespectt
ooneormor
ecr
it
eri
a.
Thet
otalv
ari
ati
onpr
esenti
nasetofdat
amaybepar
ti
ti
onedi
ntoanumberofnon-
over
lappi
ng
component
saspert
henat
ureoft
hecl
assi
fi
cat
ion.Thesy
stemat
icpr
ocedur
etoachi
evet
hisi
s
cal
l
edAnal
ysi
sofVar
iance(
ANOVA)
.Wi
tht
hehel
pofsuchapar
ti
ti
oni
ng,
somet
est
ingofhy
pot
hesi
s

9
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
maybeper
for
med.
I
nit
ial
l
y,Anal
ysi
sofVar
iance(
ANOVA)hadbeenempl
oyedonl
yfort
heexper
iment
aldat
afr
om t
he
Randomi
zedDesi
gnsbutl
atert
heyhav
ebeenusedf
oranal
yzi
ngsur
veyandsecondar
ydat
afr
om t
he
Descr
ipt
iveResear
ch.
Anal
ysi
sofVar
iancemayal
sobev
isual
i
zedasat
echni
quet
oexami
neadependencer
elat
ionshi
p
wher
ether
esponse(
dependence)v
ari
abl
eismet
ri
c(measur
edoni
nter
valorr
ati
oscal
e)andt
he
f
act
ors(
independentv
ari
abl
es)ar
ecat
egor
ical
innat
urewi
thanumberofcat
egor
iesmor
ethant
wo.

Exampl
eofANOVA:
Vent
urai
sanFMCGcompany
,sel
l
ingar
angeofpr
oduct
s.I
tsout
let
shav
ebeenspr
eadov
ert
he
ent
ir
est
ate.Foradmi
nist
rat
iveandpl
anni
ngpur
pose,Vent
urahassub-
div
idedt
hest
atei
ntof
our
geogr
aphi
cal
-r
egi
ons(
Nor
ther
n,East
ern,West
ernand Sout
her
n).Random sampl
edat
aofsal
es
col
l
ect
edf
rom di
ff
erentout
let
sspr
eadov
ert
hef
ourgeogr
aphi
cal
regi
ons.
Var
iat
ion,bei
ngaf
undament
alchar
act
eri
sti
csofdat
a,woul
dal
way
sbepr
esent
.Her
e,t
het
otal
v
ari
ati
oni
nthesal
esmaybemeasur
edbyt
hesquar
edsum ofdev
iat
ionf
rom t
hemeansal
es.I
fwe
anal
yzet
hesour
cesofv
ari
ati
oni
nthesal
es,
int
hiscase,
wemayi
dent
if
ytwosour
ces:
 Sal
eswi
thi
nar
egi
onwoul
ddi
ff
erandt
hiswoul
dbet
ruef
oral
lfourr
egi
ons(
wit
hin-
group
v
ari
ati
ons)
 Ther
emi
ghtbei
mpactoft
her
egi
onsandmean-
sal
esoft
hef
ourr
egi
onswoul
dnotbeal
l
t
hesamei
.e.t
her
emi
ghtbev
ari
ati
onamongr
egi
ons(
bet
ween-
groupv
ari
ati
ons)
.
So,t
otalv
ari
ati
onpr
esenti
nthesampl
edat
amaybepar
ti
ti
onedi
ntot
wocomponent
s:bet
ween-
r
egi
onsandwi
thi
n-r
egi
onsandt
hei
rmagni
tudesmaybecompar
edt
odeci
dewhet
hert
her
eisa
subst
ant
ialdi
ff
erence i
nthe sal
es wi
thr
espectt
oregi
ons.I
fthe t
wo v
ari
ati
ons ar
ein cl
ose
agr
eement
,thent
her
eisnor
easont
obel
i
evet
hatsal
esar
enotsamei
nal
lfourr
egi
onsandi
fnot
t
heni
tmaybeconcl
udedt
hatt
her
eexi
stsasubst
ant
ial
dif
fer
encebet
weensomeoral
lther
egi
ons.
Her
e,i
tshoul
dbekepti
nmi
ndt
hatANOVAi
sthepar
ti
ti
oni
ngofv
ari
ati
onaspert
heassi
gnabl
e
causesandr
andom componentandbyt
hispar
ti
ti
oni
ngANOVAt
echni
quemaybeusedasamet
hod
f
ort
est
ingsi
gni
fi
canceofdi
ff
erenceamongmeans(
mor
ethant
wo)
.
Ty
pesofAnal
ysi
sofVar
iance(
ANOVA)
I
fthev
aluesoft
her
esponsev
ari
abl
ehav
ebeenaf
fect
edbyonl
yonef
act
or(
dif
fer
entcat
egor
iesof
si
ngl
efact
or)
,thent
her
ewi
l
lbeonl
yoneassi
gnabl
ereasonbywhi
chdat
aissub-
div
ided,t
hent
he
cor
respondi
nganal
ysi
swi
l
lbeknownasOne-
WayAnal
ysi
sofVar
iance.Theexampl
e(Vent
uraSal
es)
comesi
nthi
scat
egor
y.Ot
herexampl
esmaybe:exami
ningt
hedi
ff
erencei
nanal
yti
cal
-apt
it
ude
amongst
udent
sofv
ari
oussubj
ect
-st
reams(
li
keengi
neer
inggr
aduat
es,managementgr
aduat
es,
st
ati
sti
csgr
aduat
es)
;impactofdi
ff
erentmodesofadv
ert
isement
sonbr
and-
accept
anceofconsumer

10
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
dur
abl
eset
c.
Ont
heot
herhand,
ifweconsi
dert
heef
fectofmor
ethanoneassi
gnabl
ecause(
dif
fer
entcat
egor
ies
ofmul
ti
plef
act
ors)ont
her
esponsev
ari
abl
ethent
hecor
respondi
nganal
ysi
sisknownasN-
Way
ANOVA(
N>=2)
.In par
ti
cul
ar,i
fthe i
mpactoft
wo f
act
ors(
hav
ing mul
ti
ple cat
egor
ies)been
consi
der
ed on t
hedependent(
response)v
ari
abl
ethen t
hati
sknown asTwo-
WayANOVA.For
exampl
e:i
ntheVent
uraSal
es,i
fal
ongwi
thgeogr
aphi
cal
-r
egi
ons(
Nor
ther
n,East
ern,West
ernand
Sout
her
n),one mor
efact
or‘
type ofout
let
’(Rur
aland Ur
ban)has been consi
der
ed t
hen t
he
cor
respondi
ng anal
ysi
s wi
l
lbe Two-
WayANOVA.Mor
e exampl
es:exami
ning t
he di
ff
erence i
n
anal
yti
cal
-apt
it
ude among st
udent
s ofv
ari
ous subj
ect
-st
reams and geogr
aphi
call
ocat
ions;t
he
i
mpactofdi
ff
erentmodesofadv
ert
isement
sandoccupat
ionsonbr
and-
accept
anceofconsumer
dur
abl
eset
c.
Two-
WayANOVAmaybef
urt
hercl
assi
fi
edi
ntot
wocat
egor
ies:
 Two-
WayANOVAwi
thoneobser
vat
ionpercel
l
:ther
ewi
l
lbeonl
yoneobser
vat
ioni
neach
cel
l(combi
nat
ion)
.Suppose,wehav
etwof
act
orsA(
hav
ingm cat
egor
ies)andB(
hav
ingn
cat
egor
ies)
,So,t
her
ewi
l
lbeN=m*
ntot
alobser
vat
ionswi
thoneobser
vat
ion(
dat
a-poi
nt)i
n
eachof(
AiBj)cel
l(combi
nat
ion)
,i=1,2,…….
,m andj
=1,2,….
.n.Her
e,t
heef
fectoft
het
wo
f
act
orsmaybeexami
ned.
 Two-
WayANOVAwi
thmul
ti
pleobser
vat
ionspercel
l
:ther
ewi
l
lbemul
ti
pleobser
vat
ionsi
n
eachcel
l(combi
nat
ion)
.Her
e,al
ongwi
tht
heef
fectoft
wof
act
ors,t
hei
rint
eract
ionef
fect
mayal
sobeexami
ned.I
nter
act
ionef
fectoccur
swhent
hei
mpactofonef
act
or(
assi
gnabl
e
cause)depends on t
he cat
egor
yofot
herassi
gnabl
e cause (
fact
or)and so on.For
exami
ningi
nter
act
ion-
eff
ecti
tisnecessar
ythateachcel
l(combi
nat
ion)shoul
dhav
emor
e
t
hanoneobser
vat
ionssoi
tmaynotbepossi
blei
ntheear
li
erTwo-
WayANOVAwi
thone
obser
vat
ionpercel
l
.
Concept
ual
Backgr
ound:
Thef
undament
alconceptbehi
ndt
heAnal
ysi
sofVar
iancei
s“Li
nearModel
”.
X1,
X2,
……….
Xnar
eobser
vabl
equant
it
ies.Her
e,al
lthev
aluescanbeexpr
essedas:
Xi=µi+ei
Wher
eµii
sthet
ruev
aluewhi
chi
sbecauseofsomeassi
gnabl
ecausesandeii
stheer
rort
erm whi
ch
i
sbecauseofr
andom causes.Her
e,i
thasbeenassumedt
hatal
ler
rort
ermseiar
eindependent
di
str
ibut
ednor
mal
var
iat
ewi
thmeanzer
oandcommonv
ari σe2)
ance( .
Fur
ther
,tr
uev
alueµicanbeassumedt
obeconsi
stofal
i
nearf
unct
ionoft
1,t
2,
…….
tk,knownas
“
eff
ect
s”.
I
final
i
nearmodel
,al
lef
fect
st’
jsar
eunknownconst
ant
s(par
amet
ers)
,thent
hatl
i
nearmodeli
s

11
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
knownas“
fi
xed-
eff
ectmodel
”.Ot
her
wise,i
fef
fect
st’
jsar
erandom v
ari
abl
est
hent
hatmodeli
s
knownas“
random-
eff
ectmodel
”.
One-
WayAnal
ysi
sofVar
iance:
We hav
e n obser
vat
ions (
Xij)
,di
vi
ded i
nto k gr
oups,A1,A2,
…….
Ak,wi
th each gr
oup hav
ing
njobser
vat
ions.
Her
e,t
hepr
oposedf
ixed-
eff
ectl
i
nearmodel
is:
Xij=µi+eij
t
h
Wher
eµii
sthemeanoft
hei gr
oup.
Gener
alef
fect(
grandmean)
:µ=Σ(
ni.µi)
/n
t
h
andaddi
ti
onal
eff
ectoft
hei gr
oupov
ert
hegener
alef
fect
:αi=µi–µ.
So,
thel
i
nearmodel
becomes:
Xij=µ+αi+eij
wi
thΣi(
niαi)=0
Thel
east
-squar
eest
imat
esofµandαimaybedet
ermi
nedbymi
nimi
zi
nger
rorsum ofsquar
e
ΣiΣjeij2)=ΣiΣj( 2
( Xij–µ–αi)as:
t
h
X.
.(combi
nedmeanoft
hesampl
e)andXi.(
meanoft
hei gr
oupi
nthesampl
e).
So,
theest
imat
edl
i
nearmodel
becomes:
Xij=X.
.+(
Xi.
-X.
.)+(
Xij–Xi.
)
Thi
scanbef
urt
hersol
vedas:
2 2 2
ΣiΣj(
Xij–X.
.)=Σini(
Xi.
-X.
.)+ΣiΣj(
Xij–Xi.
)
Tot
alSum ofSquar
e=Sum ofsquar
eduet
ogr
oup-
eff
ect+Sum ofsquar
eduet
oer
ror
or
Tot
alSum ofSquar
e=Bet
weenGr
oupSum ofsquar
e+Wi
thi
nGr
oupSum ofsquar
e
TSS=SSB+SSE
Fur
ther
,MeanSum ofSquar
emaybegi
venas:
MSB=SSB/
(k-
1)andMSE=SSE/
(n-
k),
wher
e(k-
1)i
sthedegr
eeoff
reedom (
df)f
orSSBand(
n-k)i
sthedff
orSSE.
Her
e,i
tshoul
dbenot
edt
hatSSBandSSEaddedupt
oTSSandt
hecor
respondi
ngdf
’s(
k-1)and(
n-k)
addupt
otot
aldf(
n-1)butMSBandMSEwi
l
lnotbeaddedupt
oTot
alMS.
Thi
sbypar
ti
ti
oni
ngTSSandt
otal
dfi
ntot
wocomponent
s,wemaybeabl
etot
estt
hehy
pot
hesi
s:
H0:
µ1=µ2=……….
=µk
H1:
Notal
lµ’
sar
esamei
.
e.atl
eastoneµi
sdi
ff
erentf
rom ot
her
s.
oral
ter
nat
ivel
y:
H0:
α1=α2=……….
=αk=0

12
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
H1:
Notal
lα’
sar
ezer
oi.
e.atl
eastoneαi
sdi
ff
erentf
rom zer
o.
MSEhasal
way
sbeenanunbi
asedest
i eofσe2andi
mat fH0i
str
uet
henMSBwi
l
lal
sobeanunbi
ased
est
i eofσe2.
mat
Fur
therMSB/σe2wi
l
lfol
l
ow Chi
-squar
e(χ2)di
str
ibut
ionwi
th(
k-1)dfandMSE/σe2wi
l
lfol
l
ow Chi
-
squar
e(χ2)di
str
ibut
ionwi
th(
n-k)df woχ2di
.Theset str
ibut
ionsar
eindependentsot
her
ati
ooft
wo
Chi
-squar
e(χ2)v
ari
ateF=MSB/
MSEwi
l
lfol
l
owv
ari
ance-
rat
iodi
str
ibut
ion(
Fdi
str
ibut
ion)wi
th(
k-1)
,(n
-
k)df
.
Her
e,t
het
est
-st
ati
sti
cFi
sar
ight
-t
ail
edt
est(
one-
tai
l
edTest
).Accor
dingl
y,p-
val
uemaybeest
imat
ed
t
odeci
deaboutr
eject
/notabl
etor
ejectoft
henul
lhy
pot
hesi
sH0.
I
fisH0r
eject
edi
.e.al
lµ’
sar
enotsamet
henr
eject
ingt
henul
lhy
pot
hesi
sdoesnoti
nfor
m whi
ch
gr
oup-
meansar
edi
ff
erentf
rom ot
her
s,So,Post
-HocAnal
ysi
sist
obeper
for
medt
oident
if
ywhi
ch
gr
oup-
meansar
esi
gni
fi
cant
lydi
ff
erentf
rom ot
her
s.PostHocTesti
sint
hef
orm ofmul
ti
ple
compar
isonbyt
est
ingequal
i
tyoft
wogr
oup-
means(
twoatat
ime)i
.e.H0:µp=µqbyusi
ngt
wo-
group
i
ndependentsampl
est
estorbycompar
ingt
hedi
ff
erencebet
weensampl
emeans(
twoatat
ime)
wi
tht
hel
eastsi
gni
fi
cancedi
ff
erence(
LSD)
/cr
it
ical
dif
fer
ence(
CD)
1/
2
=t
er
ror
- *
df MSE/(
1/np+1/
nq)
I
fobser
veddi
ff
erencebet
weent
womeansi
sgr
eat
ert
hant
heLSD/
CDt
hent
hecor
respondi
ngNul
l
hy
pot
hesi
sisr
eject
edatal
phal
evel
ofsi
gni
fi
cance.
Assumpt
ionsf
orANOVA
Thoughi
thasbeendi
scussedi
ntheconcept
ualpar
tjustt
orei
ter
atei
tshoul
dbeensur
edt
hatt
he
f
oll
owi
ngassumpt
ionsmustbef
ulf
il
led:
1.Thepopul
ati
onsf
rom wher
esampl
eshav
ebeendr
awnshoul
dfol
l
owanor
mal
dist
ri
but
ion.
2.Thesampl
eshav
ebeensel
ect
edr
andoml
yandi
ndependent
ly.
3.Eachgr
oupshoul
dhav
ecommonv
ari
ancei
.e.shoul
dbehomoscedast
ici
.e.t
hev
ari
abi
l
ityi
nthe
dependentv
ari
abl
eval
ueswi
thi
ndi
ff
erentgr
oupsi
sequal
.
I
tshoul
dbenot
edt
hatt
heLi
nearModelusedi
nANOVAi
snotaf
fect
edbymi
nordev
iat
ionsi
nthe
assumpt
ionsespeci
all
yift
hesampl
eisl
arge.
The Nor
mal
i
tyAssumpt
ion maybe checked usi
ng Test
s ofNor
mal
i
ty:Shapi
ro-
Wil
k Testand
Kol
mogor
ov-
Smi
rnovTestwi
thLi
l
li
efor
sSi
gni
fi
canceCor
rect
ion.Her
e,Nor
malPr
obabi
l
ityPl
ots(
P-P
Pl
otsandQ-
Q Pl
ots)mayal
sobeusedf
orchecki
ngnor
mal
i
tyassumpt
ion.Theassumpt
ionof
Equal
i
tyofVar
iances(
homoscedast
ici
ty)maybecheckedusi
ngdi
ff
erentTest
sofHomogenei
tyof
Var
iances(
Lev
eneTest
,Bar
tl
ett
’st
est
,Br
own–For
syt
het
estet
c.)
.

ANOVAv
sT-
test
:
Weempl
oyt
wo-
independentsampl
eT-
testt
oex
ami
newhet
hert
her
eexi
stsasi
gni
fi
cantdi
ff
erencei
n

13
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
t
hemeansoft
wocat
egor
iesi
.e.t
het
wosampl
eshav
ecomef
rom t
hesameordi
ff
erentpopul
ati
ons.
Theext
ensi
ont
oitmaybeappl
i
edt
oper
for
m mul
ti
pleT-
test
s(byt
aki
ngt
woatat
ime)t
oexami
ne
t
hesi
gni
fi
canceoft
hedi
ff
erencei
nthemeansofk-
sampl
esi
npl
aceofANOVA.I
fthi
sisat
tempt
ed,
t
hent
heer
ror
sinv
olv
edi
nthet
est
ingofhy
pot
hesi
s(t
ypeIandt
ypeI
Ier
ror
)can’
tbeest
imat
ed
cor
rect
lyandt
hev
alueoft
ypeIer
rorwi
l
lbemuchmor
ethanal
pha(
signi
fi
cancel
evel
).So,i
nthi
s
si
tuat
ion,
ANOVAi
sal
way
spr
efer
redov
ermul
ti
plei
ntendentsampl
esT-
test
s.
Li
ke,i
nourexampl
ewehav
efourcat
egor
iesoft
her
egi
onsNor
ther
n(N)
,East
ern(
E),West
ern(
W)
andSout
her
n(S)
.Ifwewantt
ocompar
ethepopul
ati
onmeansbyusi
ngt
wo-
independentsampl
eT-
t
esti
.e.byt
aki
ngt
wocat
egor
ies(
groups)atat
ime.Wehav
etomake4C2=6numberofcompar
isons
i
.e.si
xindependentsampl
est
est
shav
etoper
for
med(
Test
scompar
ingNwi
thS,
Nwi
thE,
Nwi
thW,
Ewi
thW,Ewi
thSandW wi
thS)
,Supposewear
eusi
ng5% l
evelofsi
gni
fi
cancef
ort
henul
l
6
hy
pot
hesi
sbasedonsi
xindi
vi
dualT-
test
s,t
hent
ypeIer
rorwi
l
lbe=1-(
0.95)=1-
0.735=0.
265i
.e.
26.
5%.
Q.
5 Whati
schi
-squar
e(X2)di
str
ibut
ion?Whatar
etheusesofchi
-squar
e(X2)di
str
ibut
ion?
Answer
:
Setupt
hehy
pot
hesi
sforChi
-Squar
egoodnessoff
itt
est
:
A.Nul
lhy
pot
hesi
s:
I
nChi
-Squar
egoodnessoff
itt
est
,thenul
lhy
pot
hesi
sassumest
hatt
her
eisnosi
gni
fi
cantdi
ff
erence
bet
weent
heobser
vedandt
heexpect
edv
alue.
B.Al
ter
nat
ivehy
pot
hesi
s:
I
nChi
-Squar
egoodnessoff
itt
est
,theal
ter
nat
ivehy
pot
hesi
sassumest
hatt
her
eisasi
gni
fi
cant
di
ff
erencebet
weent
heobser
vedandt
heexpect
edv
alue.
Comput
ethev
alueofChi
-Squar
egoodnessoff
itt
estusi
ngt
hef
oll
owi
ngf
ormul
a:

Wher
e, =Chi
-Squar
egoodnessoff
itt
estO=obser
vedv
alueE=expect
edv
alue
Degr
eeoff
reedom:
I
nChi
-Squar
egoodnessoff
itt
est
,thedegr
eeoff
reedom dependsont
hedi
str
ibut
ionoft
hesampl
e.
Thef
oll
owi
ngt
abl
eshowst
hedi
str
ibut
ionandanassoci
ateddegr
eeoff
reedom:
Ty
peofdi
str
ibut
ion Noofconst
rai
nts Degr
eeoff
reedom
Bi
nomi
nal
dist
ri
but
ion 1 n-
1
Poi
ssondi
str
ibut
ion 2 n-
2
Nor
mal
dist
ri
but
ion 3 n-
3
Hy
pot
hesi
stest
ing:

14
 Cour
se:
Educat
ional
Stat
ist
ics(
8614)
Semest
er:
Aut
umn,
2021
Hy
pot
hesi
stest
ingi
nChi
-Squar
egoodnessoff
itt
esti
sthesameasi
not
hert
est
s,l
i
ket
-t
est
,ANOVA,
et
c.Thecal
cul
atedv
alueofChi
-Squar
egoodnessoff
itt
esti
scompar
edwi
tht
het
abl
eval
ue.I
fthe
cal
cul
atedv
alueofChi
-Squar
egoodnessoff
itt
esti
sgr
eat
ert
hant
het
abl
eval
ue,wewi
l
lrej
ectt
he
nul
lhy
pot
hesi
sandconcl
udet
hatt
her
eisasi
gni
fi
cantdi
ff
erencebet
weent
heobser
vedandt
he
expect
edf
requency
.Ift
hecal
cul
atedv
alueofChi
-Squar
egoodnessoff
itt
esti
slesst
hant
het
abl
e
v
alue,
wewi
l
lacceptt
henul
lhy
pot
hesi
sandconcl
udet
hatt
her
eisnosi
gni
fi
cantdi
ff
erencebet
ween
t
heobser
vedandexpect
edv
alue.
Chi
-Squar
egoodnessoff
itt
esti
sanon-
par
amet
ri
ctestt
hati
susedt
ofi
ndouthow t
heobser
ved
v
alueofagi
venphenomenoni
ssi
gni
fi
cant
lydi
ff
erentf
rom t
heexpect
edv
alue. I
nChi
-Squar
e
goodnessoff
itt
est
,thet
erm goodnessoff
iti
susedt
ocompar
etheobser
vedsampl
edi
str
ibut
ion
wi
tht
heexpect
edpr
obabi
l
itydi
str
ibut
ion.Chi
-Squar
egoodnessoff
itt
estdet
ermi
neshow wel
l
t
heor
eti
caldi
str
ibut
ion(
suchasnor
mal
,bi
nomi
al,orPoi
sson)f
it
stheempi
ri
caldi
str
ibut
ion.I
nChi
-
Squar
egoodnessoff
itt
est
,sampl
edat
aisdi
vi
dedi
ntoi
nter
val
s.Thent
henumber
sofpoi
ntst
hatf
all
i
ntot
hei
nter
val
arecompar
ed,
wit
htheexpect
ednumber
sofpoi
ntsi
neachi
nter
val
.
Weuset
hechi
-squar
etestt
otestt
hev
ali
dit
yofadi
str
ibut
ionassumedf
orar
andom phenomenon.
Thet
estev
aluat
est
henul
lhy
pot
hesesH0(
thatt
hedat
aar
egov
ernedbyt
heassumeddi
str
ibut
ion)
agai
nstt
heal
ter
nat
ive(
thatt
hedat
aar
enotdr
awnf
rom t
heassumeddi
str
ibut
ion)
.
Letp1,
p2,
..
.,
pkdenot
ethepr
obabi
l
iti
eshy
pot
hesi
zedf
orkpossi
bleout
comes.I
nni
ndependentt
ri
als,
wel
etY1,
Y2,
..
.,
Ykdenot
etheobser
vedcount
sofeachout
comewhi
char
etobecompar
edt
othe
expect
edcount
snp1,
np2,
..
.,
npk.Thechi
-squar
etestst
ati
sti
cisqk-1=

=(
Y1-np1)
²+(
Y2-np2)
²+.
..+(
Yk-npk)
²
-
--
--
--
--
---
--
--
--
-- -
--
--
--
-
np1 np2 npk

Rej
ectH0i
fthi
sval
ueexceedst
heupper cr
it
icalv
alueoft
he (
k-1)di
str
ibut
ion,
wher
e i
sthe
desi
redl
evel
ofsi
gni
fi
cance.

15