Testing Statistical Hypothesis:

A hypothesis is a statement whether something is true

In this chapterwe willlearnhowtotest hypothesis claims aboutunknown
population parameters such as mean proportions variance correlation and
independence

Examples of claims
Theaverageheights ofmaleadults in westbank is175an
Lessthan 5 ofthefemaleadults in westbank smoke cigarettes
There's norelation between Taw hi average andthe university GPA
ji
Unemployment rate is more than 1

Relation Negation

E
7 C
7
f s

In testingstatistical hypothesis about parameterof the population you

have to testtwo hypothesises
Ho null hypothesis
HalHa alternative hypothesis
 In testing statistical hypothesis there's always chance of making
errors and one neverget 100 accurateresults unless we are studying
the wholepopulation which is impossible in mostcases

Actual situation
the is false

É don'trejectto correct Type II error

decision 1B errol
rejectits Trisector Lefton

Type I error x error

Mistake of rejecting Howhile it's true

Type II error p error

Mistake of acceptingIto while it's false

Thesignificance level x P Type I error

PCrejectingHo Hoistrue
usually x is chosentobesmallsuchas no 4 0.01 a 0.02 a 0.05 a o lo

B a symbolusedtodenotetheprobabilityof type II error

B P Type II error P lacceptingHo Ho is false

is Thepoweroftesting 1 13 PlrejectingHo I Hoisfalse

Choice of Ho
Howill correspond to theoriginalclaim
If theclaimincludes thecondition ofnochange or differences equalitysign
otherwise
Ho is the negation oftheclaim
 Example
Claim Ho Ha
M 775 ME75 M775
0.05 p 0.05 P 0.05
p 7 0.8 p 7 0.8 P C 0.8
M Me M EMa M Mz
e Ci P re P Sf

Theessentialsteps fortestinghypothesis
1 Identifythe claim and express it in symbolic form
2 Expressthenegation oftheclaim in symbolicform
3 Outofthetwoexpressions above thenull hypothesis Ito is theone withcondition
of equality and He is theother
4 Basedonthe seriousness oftype I error choose x
5 Determine which estimator is appropriateandwhatis theapproximate samplingdistribution
6 Calculatetheteststatistic this isthe statisticwhosevaluedepends onthesample
dataand isusedto test whetherwe accept or rejectHo
7 Findthecritical value s which are the values thatdependon the distribution
oftheteststatistic
AUsingvalues in 6and7 eitherwerejectto or fail to reject accepts Ho
9 Usinganswer in 8 we givetheconclusion abouttheclaim

Testingclaimsaboutthe population mean

The population mean is denotedby m
The sample mean is denoted I
by
 To test claims such as
re 100 y 80 M 15 at significant
level x depending on a random sample ofsize n from a population X the
teststatistic and criticalvalues dependon whetherthe population variance
is known andwe havethefollowing cases

case11 Two tails test

Ho Melo vs He M Mo
It is known Test statistic
a o Z
Iff
Critical values I Zaz

i
b If o is unknown Teststatistic t
Is to ten 1
Criticalvalues I taiz

Case 12 Left tail test

Ho M Mo vs H1 MsMo
a It o is known Teststatistic Z
Iga
Criticalvalue za

gift
b If is unknown Test statistic t I Mo s tr ten 1
Str
Critical value ta

riff
ta
Case 3 Bright tail test
Ho Mello vs H1 MSM
a If o is known Test statistic Z
IG
Criticalvalue za Plz za a

plz za 1 x

Iggy
Respecting

b If o is unknown Teststatistic t IgG tn ten 1

Criticalvalue ta

Example
A manufacturers claimthattheaverage lifetime of the light bulbs produced
bytheirmachine is at least 1000hours asample of 25 bulbs produce a man

of 990hoursatsignificant level x 0.05 Doyouaccept or rejecttheirclaim

assuming
a 40
Claim M 1000 Ho 171000 HI M 1000
4 25 I 990 4 0.05
is left tail test
6 is known as I n NCtooo 169g

Test statistic Z To 99 7900 1 25

Criticalvalue ta Zoos
7 1.65
P Z C zoos o og
1.647 Zo05 1.6 45
Zz

Wedon't failto rejectHo since z is in the accepted

ayyy rejeon so weaccepttheclaim

b o is unknown andthe sampleproduce variance 900hours

is unknown 52 900
Test statistic t 99 7,000
Ijn 1.67
Criticalvalue toos
PL t to s 0.05 If n 1 24
to05 1.711
toog 1.711

Wedon'treject Ho so weacceptthe claim

Example
Salt is packedin 2kg boxes a sample of 49boxes producedan average
weight of 1.9kg and variance 0.09kg At significant level 0.01
Claim M S2
Ho
M 2 Hz u c z n left tail test
Mo 2 I 1.9 52 0.09 2 0.01 n 49
o n unknown E n t148

Teststatistic t 2.33
TIE 18.3

Critical value toot

Plt C to 1 0.01
to.o1 2.423
Wedon'treject Ho
We rejecttheclaim sothecustomerisn'tcheated
Example
Find 99 confidence interval forthe average height ofmaleadults in
WestBank it a sample of size 25 produced an average height 175 em

and variance 49
Testthe claim thattheaverage heightof maleadults is 170 an at
4 0.01 in twoways given thatthe heights are normally distributed
62ns unknown pop is normal
4 0.01 n 25 I 175 s 49
4 0 005
C I I tazz I t taiz sa
taz Plt taiz I X
PL t tan Nz
P t to.us 0.005 ns to.us 2.797

99 C I 175 12.797 175 2.797157

C 171 I 178.93

Claim M 170 Ho M 170 HI Mt 170

Method I

µ 170 99 C I 171.1 175.97

So we reject Ho and rejectthe claim

Method 2 two tail test

Test statistic t
Iff 177,170 3.5714

Critical values I tan

t n t 24
 to005 2.797

da 3 my
t taiz in therejecting region

Wereject Ho so we reject the claim

Example
Let X be NLM 4 To test Hoagainst H Mto
A 0 we
take a randomsample of size n 25 fromthis dist and observe
that I 0.28 Do we accept or reject Ho at the 10 significance
level Is 1 0 contained in a 90 C I form
62 4 n known pop is normal
1 25 I 0.28 X 0.1
Ho M 0 HI M O

Two tail test

Critical values I Zaiz I Zoos
P Z Zoos 0.05
PLZ CZoos 0.95
I Zoos I 1 645
Test statistic Z 0.7
Fyn 0.2950

t.ms
Eimacceptingregion
Weaccept fail to reject Ho
n Since we accept 1 0 at x 10
i
MIO E I N C I

1 0 E 90 C I

Example
Let X be Nn IM To test Ho Mets against Hr M 7s
2

we took a random sample of size 25 andobserved that 7 78.8

and 5 12.8 Do we accept or reject Ho at 5 significance
level
o n unknown Ent 24
5 12.8 1 25 I 78.8 4 0.05
Ho
M 79 He M 75 is two tail test

Criticalvalues tan
I
P t tan y at df 24
PL t to.org 0.025
fromthe table I tan I 2.064

Test statistic t 1 484

Iya 7923,75

2.064 L t L 484 2.064 is belongs to acceptance

region
So we accept Ho
 Testing claims about population proportion

Thepopulation proportion is denoted P

by
the sample proportion is denoted
by I In
X thenumberof successes in
the n trails

Totestclaims such as 1 0.5 at significance level

P 70.0 P70.6
x we take n large enough so that up and
ng are atleast 5 and we
use normal approximation to binomial distribution
So we use thefollowingstatistic
Teststatistic n z
typing
PP
Paf
Criticalvalues as I Zaz for two tails test
za for best tail test
one tailtest
Za for right tail test

belt tail test

TÉ
right test
tail
zaz Zaz
two tail test

Example
A manufacturer claims that lessthan 10 of the items produced byhis
machine are defective at significance level a 0.05
Test his claim if
a sample of size 1000 items had 95 defective items
Cairn P o 1
Ho P 70.1 Hz PCO I
 N 1000 n nP 1000 0.1 too 75
11000 o 9 900
the dist is approximal
ng s normal

Since Ho is P 0.1 the test is left tail test

The criticalvalue L Ea
plz za
Zoog
X 0.05
1 645
Éx
Grejectingregion

Teststatistic
Z P P P In 9
Pat

Z 0.095 O I 0.5270
10.116.911

ns Since o 52707 1.645 12722

Weaccept Ho
and we rejectthe claim

Example
Arandomsample of size 36within proportion 68 is drown from a population
At significance level x 0.02 testthe claimthattheproportion ofthe population
is 60
Claim P 0.6 Ho p o6 He p 0G
n 36 p 0.6 p 0 65 412 0.01

Two tail test Criticalvalues I Zan

P Z Zoo 0.01
 P Z C Zoo 0.99 ns Zoo 2.33

the
Test statistic Z P P 0 9797
Pat É if z

2.33 C 0.9797 4 2.33

We accept don'treject H
So we accept theclaim

up 21.6 75 and
ng 14.4 75
So n is largeenough to approximate to normaldist

Confidince interval for population proportion

If p is the proportion ofthepopulation and p is the proportion of the
random sample with size n drawnfromthe population then the I a CI
for p is

P Zai PT P Zar PF

The 11 4 100 C I for p means that

PIp Zapf f p f p Zan PIE 1 a

Example
Arandom sample of size 64 with proportion p Goto is drawn from a
population Find95 C I for p
P 0.6 4 0.05 n 64
pep 0.6 no pointestimation
 Zaz P Z Zan I L
P Z C Zo oz 0.975
Zo025 1.96

95 C I p ZarPEF P Zant
o 6 1.96 6.634T OG 11.96
t.gg
C o 48 0.723

Note

Tht're
E marginal error
2 Interval length L ZE

Determining the sample size for theestimation of proportion

Giventhe confidence intervalandthe values of p and g the sample
size that willproduce a predetermined margin of error E of the
confidence estimate for
p is
11 41 100 C I CP E P E
p ZarPIT p Zan
Length of C I ft za Igf
L
p za PIF
L 22 12
PIT

Iq 2212 n n pg p Eazy

n pg ZI
 n mustbe integer If we didn't get an integerwe round it up

Example
How large a sample with proportion 0.07 shouldbe to
give a 95 C I
for p within 0.02 of thepopulation proportion

marginal error
9 0.05 E 0.02
n Piggy

Zan PlZ 720.02s 0.025

P Z C Zo.org 0.975
Z 0.02s 1.96
N 6 07710,93 1.9632 625 22 a 626
10.0212
roundup

Dana Nabil Hamzah