HYPOTHESI

STESTI
NGBYMUREBAM
B
HYPOTHESI
S
 I
tisanideaoraproposi
ti
ont
hatcanbesubj
ect
edt
ofur
theri
nvest
igat
iont
o
t
estit
sv al
idi
ty.
 Sampledat
aisanal
ysedt
oenabl
econcl
usi
onsorgener
ali
sat
ionsont
heent
ir
e
popul
ati
on
 Therearetwot
ypesofhy
pot
hesi
s.Thesear
enul
lhy
pot
hesi
sandal
ter
nat
ive
hypothesi
s
Nul
lhy
pot
hesi
s
 Anul
lhy
pot
hesi
sisdenot
edbyH0
 Thenull
hypot
hesisisast
atementwhi
chi
mpl
iest
hatt
hei
snochangei
nthe
popul
ati
onparameter
Al
ter
nat
ivehy
pot
hesi
s
 Al
ter
nat
ivehy
pot
hesi
sisdenot
edbyH0
 Theal
ternat
ivehy
pothesi
sisast
atementi
mpl
iest
hatt
her
eisachangei
nthe
popul
ati
onparameter
 Thechangemay
bespeci
fi
cornon–speci
fi
c
 Aspeci
fi
cchangei
mpl
iesei
therani
ncr
easeoradecr
easei
nthepopul
ati
on
par
ameter
 I
fthechangei
sspeci
fi
cthehy
pot
hesi
stesti
sknownasaonet
ail
edt
est
 I
fchangei
nthepar
amet
eri
snotspeci
fi
cthet
esti
scal
ledat
wot
ail
edt
est
 Attheendofahypothesi
stestaconcl
usi
oni
smade.Thi
sconcl
usi
onmaybe
accept
anceorr
eject
ionofH0
Cr
it
icalv
alue
 Iti
stheval
uewhichf
ormt
heboundar
yoft
her
eject
ionr
egi
on(
cri
ti
cal
)andt
he
accept
anceregi
on
Testst
ati
sti
c
 I
tisav
aluecal
cul
atedf
rom sampl
edat
aandusedf
ort
est
ingpur
poses
  I
fthet
estst
ati
sti
cisi
ntheaccept
ancer
egi
onH0i
saccept
ed
 I
fthet
estst
ati
sti
cisi
nther
eject
ionr
egi
onH0i
srej
ect
ed
Lev
elofsi
gni
fi
cance
 Alevelofsigni
fi
canceisusuallyst
atedasapercentage,t
hispercent
age
deter
minest hepercentageofthedi
str
ibut
ionwhichformsther ej
ect
ionregi
on
egifthelev
elofsignifi
canceis5%thentherej
ecti
onconst i
tut
es5%oft he
di
stri
buti
on

Hy
pot
hesi
sforʯ
Hy
pot
hesi
stestf
orʯi
smadeupoft
hef
oll
owi
ngst
ages
1.Def
ini
ngt
hev
ari
abl
e
2.St
ateH0andH1

Foraonet
ail
edt
estH1i
ssucht
hatʯ>ʯ0
 H1:ʯ>ʯ0 (
uppert
ailt
est
)
 H1:ʯ<ʯ0 (
lowert
ailt
est
)
Forat
wot
ail
edt
est
H1:ʯ ‡ʯ0
3.St
atet
hedi
str
ibut
ionoft
het
estst
ati
sti
cassumi
ngH0i
str
ue.
 Whent
est
ingf
orʯt
het
estst
ati
sti
cisȾsowest
atet
hedi
str
ibut
ionofȾ
4.St
atet
het
ypeoft
het
estandl
evelofsi
gni
fi
cance
5.Fi
ndt
hecr
it
icalv
alue
6.Fi
ndt
het
estst
ati
stc
 St
andardi
zet
heteststat
ist
icanddr
awadi
agr
am showi
ngposi
ti
onsoft
est
st
ati
sti
candcri
ti
calval
ue
7.Concl
usi
on
 Theconcl
usi
onconsi
stoft
wopar
ts,
stat
ist
icalandcont
ext
ual
1.St
ati
sti
calconcl
usi
on
 Rej
ectH0ordonotr
ejectH0dependi
ngont
heposi
ti
onoft
het
estst
ati
sti
c
onthedi
agram
 2.Cont
ext
ual
concl
usi
on
 Rel
atet
hest
ati
sti
calconcl
usi
ont
othecont
extoft
hequest
ion

EXAMPLES
1.AnIQtesti
sdevelopedsuchthatthemeanquot i
enti
s100andstandard
devi
ati
onis12.I
tisgiventoarandom sampleof50chil
dreni
n1area.The
aver
agemar kwas105.Doesthisprovi
deevidenceatt
he5%level
,thatchi
ldr
en
fr
om thi
sareaaregeneral
lymoreint
ell
igent

Sol
uti
on
Letxbet
heI
Qofachi
ld
H0:ʯ=100
H1:ʯ>100
144
Ⱦ ͠ N(
100 )
50
Onet
ail
edt
estuppert
ail
Testat5%l
evelofsi
gni
fi
canceandr
ejectHoi
fZcal>1,
645 cr
it
icalr
egi
on1-5%on
tabl
eor
P(
z<z)=0,
95
Φ(
Z)=0,
95
Z=1,
645
Testst
ati
sti
c

Ⱦ-ʯ
Zcal =
δ∕ n
105-100
=
12∕ 50
=2,946
2,
946>1,
645 i
nequal
it
yforr
eject
ion
Concl
usi
on
WerejectH0andconcl
udet
hatt
hechi
ldr
enf
rom t
hisar
eaar
egener
all
ymor
e
i
ntel
li
gent.
24
2.Asampleof40observat
ionsfr
om anormaldist
ri
buti
on,
x,gave∑x24,Ⱦ=
40
2
and∑x =596.Perf
orm atwotail
edtestatt
he5%l ev
elt
odetermi
newhet
her
themeanofthedi
str
ibut
ionis0
Sol
uti
on

1 (∑x)
2
δ= [
∑x²– ]
n-1 n

1 (24)
²
= [596– ]
39 40
=3,
862
H0:
ʯ =0
H1:ʯ ‡0

Ⱦ ͠ N( 8622∕40)
03,
Twot
ail
edt
est
Testat5%l
evelandr
ejectH0i
f│ Zcal│ <1,
960 (
1-5%∕2)

0,
6-0
Zcal =
826∕ 40
3,
=0,
983
I
nequal
it
yforr
eject
ion
-
1,960<0,
983<1,
960
Concl
usi
on
Wedonotr
ejectH0andconcl
udet
hatt
hemeanoft
hedi
str
ibut
ioni
s0
3.Ami ll
ingcompanyclai
mst hatabagofflouri
tpackswei
ghs10kgeachon
average.Arandom sampleof50bagsisexami nedandt
hemass,xkg,
oft
he
contentsofeachbagisrecorded.I
tisf
oundthat∑( x–10)=-
12,3and∑(x–
2
10)=37, 7
a.Est
imat
ethepopul
ati
onmeanandv
ari
anceoft
hemassoft
hecont
ent
s
ofabag
 b.Test
,atthe10%levelofsi
gni
fi
cance,whethert
hemill
ingcompanyi
s
over
stat
ingtheaver
agemassoft hecontentsi
neachbag.
Sol
uti
on
1)N=50

ˆ
∑(
x-10)
ʯ =Ⱦ = +10
n
-
12,
3
= +10
50
=9,
754

1 2
[
∑(x-a)
]²
Δ2 = [∑(
x-a)- ]
n-1 n
1 (
[-12,3)
]²
= [37,
7- ]
49 50

=0,
708
2)H0:ʯ =10
H1:ʯ<10
0,
708
Ⱦ ͠ N[
10 ]i
fʯ=10
50
0net
ail
edt
estl
owert
ail
Testat10%l
evel
ofsi
gni
fi
cantandr
ejectH0i
fZcall<-Zcrit
Zcal<-
1,282

9,
754-
10
ZCal =
0,
708
50
=-
2,067
I
nequal
it
yforr
eject
ion
 -
2,067<-
1,282
Concl
usi
on
Werej
ectH0andconcl
udet
hatt
hecompanyi
sov
erst
ati
ngt
heav
eragemassof
t
hecontent
sineachbag

TESTI
NGʯWHENTHE
POPULLATI SNORMALδ2
ONI
UNKNOWNANDSAMPLESI
ZEN <
30
 I
fasmallsampleisdrawnf
rom anormalpopul
ati
onwit
hunknownv
ari
ance,
t
hehypothesi
stesti
sconduct
edusingtdist
ri
buti
on.
 Thet
estst
ati
sti
cisTwher
e
Ⱦ -
ʯ
T= andT ͠ t
(n-
1)
δ∕ n

EXAMPLES
1.Thecholesterol
level
sof8womenweremeasuredwi t
hthefollowi
ngresul
ts3,
12, 81, 51, 61, 72,81,93,3makinganynecessaryassumptionst
est,at
the5%lev el
,whethert
hesampl
ehasbeendrawnfrom adistr
ibuti
onwit
hmean
chol
est
er ol3,1
Sol
uti
on
Letxbet
hechol
est
eroll
evel
sof
N=8
∑x=18,
7

∑x2=47,
49

18,
7
Ⱦ=
8
=2,
3375
 1 (18,7)
²
Δ= [
47,
49-
7 8
=0,
7347
H0:ʯ =3,
1
H1:ʯ ‡3,
1
Twot
ail
edt
est
Ⱦ-
ʯ
Testst
ati
sti
cisTwher
eT= andT ͠ t
(n-
1)
δ∕ n
Testat5%l
evelandr
ejectH0i
f│ Tcal│ <2,
3375

2,
3375-
3,1
TCAL =
0,
7347∕ 8
=-
2,935
I
NEQUALI
TYFORREJECTI
ON
-
2,935<-
2,363
CONCLUSI
ON
Werej
ectH0andconcl
udethatt
her
eisnoevi
dencethatthesampl
ehasbeen
dr
awnfrom adi
str
ibut
ionwi
thmeanchol
est
eroll
evel3,1
Assumpt
ion
Weassumedt
hatt
hesampl
ewasdr
awnf
rom anor
malpopul
ati
on
2.Arandom sampl
eof8obser
vat
ionsofanor
malv
ari
abl
egav
e∑x=36,
5and∑
(
x-Ⱦ)²=0,
74
Testatt
he5%l evelt
hehypothesi
st hatthemeanoft
hedist
ri
but
ionis4,
3
agai
nsttheal
ternat
ivehy
pothesisthatthemeanisgr
eat
erthan4,3
Sol
uti
on
N=8
∑x=36,
5
∑(
x-Ⱦ)
²=0,
74
36,
5
Ⱦ=
8
 =4,
5625

1 (∑x-Ⱦ)
2
δ= ∑x²–
[ ]
n-1 n

=0,
3251
H0:ʯ =4,
3
H1:ʯ >4,
3
Onet
ail
edt
estuppert
ail
Ⱦ-
ʯ
Testst
ati
sti
cisTwher
eT= andT ͠ t
(n-
1)
δ∕ n
P(
T<t
)=0,
95wherT ͠ t
7

=1,
895
Testat5%l
evelandr
ejectH0i
fTCAL>1,
895
Testst
ati
sti
c

4,
5625-
4,3
TCAL =
0,
3251∕ 8
=2,
284
I
nequal
it
yforr
eject
ion
2,
284>1,
895
Concl
usi
on
Wer
ejectH0andconcl
udet
hatwehav
efoundev
idencet
hatmeani
sgr
eat
ert
han4,
3
3.I
tisthoughtthatanormalpopulati
onhasmean1,6.Ar
andom sampleof10
observat
ionsgivest
hemeanof1, 49andst
andar
ddevi
ati
on0,3.Doesthi
s
provi
deev i
denceatthe5%levelthatt
hepopul
ati
onmeanisl
essthan1,6
Sol
uti
on
N=10
Ⱦ=1,
49
ʯ=1,
6
S=0,
3
n
δ²
= ×s²
n-1
10
= ×0,
3²
9
=0,
1
H0:ʯ =1,
6
H1:ʯ <1,
6
Ⱦ-
ʯ
Onet
ail
edt
estl
owert
ailTestst
ati
sti
cisTwher
eT= andT ͠ t
(n-
1)
δ∕ n
P(
T<t
)=0,
05wherT ͠ t
9

=-
1,833
Testat5%l
evelandr
ejectH0i
fTCAL<-
1,833

1,
49-
1,6
TCAL =
0,
1 ∕ 10
=-
1,1

-
1,1>-
1,833
Wedonotr
ejectH0andconcl
udet
hatt
her
eisev
idencet
hatt
hepopul
ati
onmeani
s
l
esst
han1,6

TESTINGABINOMIALPROPORTI
ON
P(LARGESAMPLES)
 I
fnp>5andnq>5,
nor
maldi
str
ibut
ioni
susedasanappr
oxi
mat
iont
obi
nomi
al
di
str
ibut
ion.
 Whent
est
ingf
orPi
nthi
scaseweuset
heZ-
dist
ri
but
ion.
 Thet
estst
ati
sti
cisxwher
eX͠ Bi
n(n;
p)
 Ⱦ͠ N(
np;
npq)
 x-np
 Thestandardi
zedtestst
ati
sti
cisZCALwher
eZCAL= wher
exi
sthenumber
npq
ofobservedsuccessesi
nthesample

EXAMPLES
1)I
tist houghtthatthepropor t
ionofdefect
ivepenspr oducedbya
parti
cularfactoryis0,15.Ar andom sampleof200pensi sinspect
edand
foundt ocontain38def ecti
vepens.Doest hisprovi
deev i
denceatt
he5%
l
ev elthatt
hef actoryi
spr oducingmoredef ect
ivepensthanexpected
Sol
uti
on
Letxbet
henumberofdef
ect
ivepens
H0:P=0,
15
H1:P>0,
15
X͠ Bi
n(2000,
15)i
fp=0,
15
Np>5
Nq>5
Ⱦ͠ N(
3025,
5)
Onet
ail
edt
estuppert
ail
P(
Z<z)=0,
95
=1,
645
Testat5%l
evelandr
ejectH0ZCAL>1,
645
38-30
ZCAL=
25,5

=1,
584
1,
584<1,
645
Wedonotr
ejectH0andconcludet
hatt
hei
snoev
idencet
hatt
hef
act
ory
i
spr
oduci
ngmor edefect
ive
3.Acoi
nistossed150ti
mesand67headswer
eobt
ained.Testat5%l
evel
whet
hert
hecoinisfai
r
Sol
uti
on
Letxbet
henumberofheads
 H0:P=0,
5
H1:P‡0,
5
X ͠ Bi
n(150 0,
5)i
fp=0,
5
Np>5
Nq>5
Ⱦ ͠ N(
7537,
5)
Twot
ail
edt
est
Testat5%l
evelandr
ejectH0i
f│ZCAL│<1,
960
67-75
ZCAL=
37,5

=-
1,306
-
1,960<-
1,306<1,
960
Wedonotr
ejectH0andconcl
udet
hati
tpr
ovi
desev
idencet
hatt
hecoi
nisf
air
4.Aninvest
igat
ionintotheownershipofmobil
ephonesamongstschoolchil
dren
wascarri
edout .Arandom sampleofsi
ze200wassel ected.Ahypot
hesi
stest
atthe5%levelwasdoneusi ngH0:P=0,65andH1:p>0,65.Findt
hecri
ti
cal
regi
onforthetest.
Sol
uti
on
Letxbet
henumberofchi
l
drenwhoownmobi
lephones
H0:p=0,
65
H1:p>0,
65
X ͠ Bi
n(200 0,
65)i
fp=0,
65
Np>5
Nq>5
Ⱦ ͠ N(
13045,
5)
Twot
ail
edt
est
Testat5%l
evelandr
ejectH0i
f│ZCAL│<1,
645
CRI
TICALREGI
ON
ZCAL >1,
645
 x-
130
>1,
645
45,
5

X >130+1,
645 45,
5

X >141,
0961316
X >142

TYPE1ANDTYPE2ERRORS
 Aty
pe1er
roroccur
swhenH0i
swr
ongl
yrej
ect
edi
erej
ect
ingH0wheni
t
i
str
ue
 Thepr
obabil
it
yofmaki
ngat
ype1er
rori
sequalt
othel
evel
of
si
gni
fi
canceofthet
est
 Atype2err
oroccur
swhenH0i
swr
ongl
yaccept
edi
eaccept
ingH0when
i
tisfal
se
 Toeval
uateprobabi
li
tyofatype2er
ror
,apar
ti
cul
arv
aluemustbest
ated
i
nthealt
ernat
ivehypothesi
s
 Thedi
str
ibut
ionoft
het
estst
ati
sti
cisst
atedaccor
dingt
othi
sval
ue
 Thepowerofat
est=1–p(
type2er
ror
)
EXAMPLES
1.Arandom v
ari
ablehasanor mal di
str
ibuti
onwithmeanʯandst andard
devi
ati
on3.Thenullhypothesisʯ=20istobet est
edagainstt
heal t
ernat
ive
hypot
hesi
sʯ>20.Usi ngar andom sampleofsize25.Iti
sdecidedthatthenul
l
hypot
hesi
swill
berejectedifthemeani sgreat
erthan21,4.
a)Cal
cul
atet
hepr
obabi
li
tyofmaki
ngat
ype1er
ror
b)Cal
cul
atet
hepr
obabi
li
tyofmaki
ngat
ype2er
rorwheni
nfactʯ=21
Sol
uti
on
N=2
Δ=3
a)H0:ʯ=20
H1:ʯ >20
Onet
ail
edt
estuppert
ail
 3²
Ⱦ ͠ N(
20 )i
fʯ=20
25
Cr
it
icalr
egi
on:Ⱦ>21,
4
P(
type1er
ror
)=p(
Ⱦ>21,
4)
21,
4-20
=p(
Z> )
9
25
=P(
Z>2,
333)
=1–Φ(
2,333)
=1–0,
9902
=0,
0098
b)H0:ʯ=20
H1:ʯ=21
3²
Ⱦ ͠ N(
21 )i
fʯ=21
25
H0i
srej
ect
edf
orȾ>21,
4henceH0i
saccept
edf
orȾ<21,
4
P(
type2er
ror
)=p(
Ⱦ<21,
4)
21,
4-21
=p(
Z> )
9
25
=p(
Z<0,
667)
=Φ(
0,667)
=0,
7477
2.I
tisgiventhatxfollowsanor maldist
ri
buti
onwi t
hpar ametersʯand25.Iti
s
desi
redtot estanullhypothesi
sʯ=20agai nstanalternat
ivehy
pothesi
sʯ<20
wit
hthepr obabil
i
tyofat ype1er r
orbeing5%.Ar andom sampleof100
observ
ationsofxi staken
a.Fi
ndt
heaccept
anceandr
eject
ionr
egi
ons
b.Forthecaseʯ=19,
findt
hepr
obabi
li
tyofat
ype2er
rorandt
hepoweroft
he
test
SOLUTI
ONS
a.H0:ʯ=20
H1:ʯ<20
Onet
ail
edt
estl
owert
ail
25
Ⱦ ͠ N(
20 )i
fʯ=20
100
Testat5%l
evelandr
ejectH0i
fZCAL<-
1,645
CRI
TICALREGI
ON

ZCAL<-
1,645

Ⱦ-
20
<-
1,645
25
100
5
Ⱦ <20–1,
645×
10
Ⱦ <19,
1775
Accept
ancer
egi
on:Ⱦ>19,
1775
b.H0:ʯ =20
H1:ʯ=19
25
Ⱦ ͠ N(
19 )i
fʯ=19
100
H0i
saccept
edi
fȾ>19,
1775
P(
type2er
ror
)=P(
Ⱦ>19,
1775)whenʯ=19
19,1775-19
=p(
Z> )
25
100
=p(
z>0,
355)
=1–Φ(
0,355)
=1–0,
6385
=0,
3613
 Poweroft
het
est=1-0,
3613
=0,
6387

TESTFORABINOMIALPROPORTI
ONP
(SMALLSAMPLES)
 Incaseswhereni
ssmal
lwecannotusenor
maldi
str
ibut
ionasan
approxi
mati
ontobi
nomi
aldi
str
ibut
ion
 Wecar
ryoutabi
nomi
alt
est

Pr
ocedur
efort
het
est
1.Def
inet
hev
ari
abl
e
2.St
ateH0andH1
3.St
atet
hedi
str
ibut
ionofxaccor
dingt
oH0
4.St
atet
het
ypeoft
het
estandt
hel
evel
ofsi
gni
fi
cance
5.St
atet
her
eject
ioncr
it
eri
on
 Rej
ectH0i x≤x)<β%f
fp( oraonet
ail
edt
ailt
est
 Rej
ectH0i X ≥x)<β%f
fp( oraonet
ail
edt
estuppert
ail
1 1
 Fora2t
ail
edt
estwer
ejectH0i
fei
t x≤x)< β%orp(
herp( X ≥x)<
2 2
β%
6.Fi
ndt
het
estst
ati
sti
c
7.St
atet
heconcl
usi
on
Exampl
es
Acert
aintypeofseedhasager minat
ionrat
eof70%.Theseedsundergoanew
t
reat
mentaf t
erwhich,9ger
mi nat
einapacketof10seeds.Testatt
he5%level
whet
herthereisanincr
easeinthegerminat
ionr
ate
Sol
uti
on
Letxbet
henumberofseedswhi
chger
minat
e
H0:p=0,
7
H1:p>0,
7
X ͠ Bi
n(10 0,
7)i
fp=0,
7
Onet
ail
edt
estuppert
ailandl
evelofsi
gni
fi
cance=5%
Butx=9
Rej
ectH0i x≥9)<5%
fp(

x≥9)=C910( 9 1
+C1010( 10 0
P( 0,7)(0,
3) 0,7) (0,
3)
=0,
149308345
=0,
149
0,
149>5%
Wedonotrej
ectH0andconcl
udet
hatt
her
eisev
idenceofani
ncr
easei
nthe
ger
minat
ionr
ate.