Dixteib wt on

UNIT-L

Sawpl Disib wt ion

Le (X, n) T.C duaon fom
epu, wth dus,
neF.The 7.V
Ccn be Coneideied co mdep
amd elutic ally dictsibuted wuh the d t ,

wnc, FG). L )be amy

un thuse TV
The
,
Y.V amd z
( X - -*)bewa
T.VA aleo a Y.V Te onet, z &,,X
C l zd the Sampiwa
A dht, s4 the possi bu merms

the S uwwptes n o on a Le

Selest a ho dup., TAndonm amm ple of a

a te Size vom a i ven po pu, and a leulate
Some t a t s tic
e.an ,P ctc,) F>w the

Sample Ce et Sexes Vahes af thise

tabu hiCTtuce Vales cbt oned mhe

Sommpls Cam be ped w the Ferm

re4 dut, tRi duit., So for med

af al Passi bu S tatstie 1
Val ues
Caled the s amph dist, cr pvobabl ty
Samph'ng that Statis trc
 2
pes
Same
(1 AM a +he Sampka d
mean he ves4e popu) trem
a h
whi ch Sompl eRe ake
Ls to
AM c Somplin dt cqual
(u
+e Vm.e aw
pppu .D
has a.
dist s eam

PDpu s.D divi dud

altotLe

rd. te SaMple size

the ahase voot of
y
(/)

Sbamdas d Error
te S cuwmpi hy dist, e
RYOY. SE
as
St otistic
o detes i ne wihtn ouich
6ancd tudeful
to ke
the pasaMmetesg VaMuec nt 2 p e cted
styumeut teat wg a

Ve pothes

Ex, SE( = /R
sF C =
2
 O M E - T C o n t t m u a h o n a

SIGNIFICANCE
(Basic Concen
4. TEST OF

4.0 Introduction: information about poDIuias.

all the
to collect
It is not easy the
characteristics of the n
to study
and also it is not possible factor and
due to time factor, cost
infinite)
population (finite
or

need sample. Sample

is a finite subset of
other constraints. Thus
we
the number of individuale
statistical individualsin a population and
size.
in a sample is called the san1ple
in our day-t0-day practical life
Sampling is quite often used
of rice, wheat or any
For example in a shop we assess the quality
other commodity by taking a handful of it from the bag and then to
decide to purchase it or not.

4.1 Parameter and Statistie:

3 h e statistical constants of the population such as mean, (4).|
variance (o), correlation coefficient (P) and proportion (P) are
called Parameters.
Statistical constants computed from the samples
corresponding to the
parameters namely mean (x ), variance
sample correlation coefticient (r) and proportion (S).
(p) etc, are callea
statisticA
ParameterS are functions of the
statistic functions of the
are population values while
sample observations. In general
population parameters are unknown
as their estimates. and sample statistics are uscu
4.2 Sampling Distribution:
8The distribution
assumed by some of all possible
statistic measured values which can
randomly drawn from the from samples
sampling distribution of the same population of size of same size" n
statistigs (DANIEL and N. is called a
Consider a
sample of size n frompopulation with N
FERREL).
this values Let us take a
population, then there are ranu
 2(,-)
K
4.3 Standard Error:
A
deviation of the sampling distribution_of
a
The standard
as S.EFor
its standard error. It is abbreviated
statistic is known as
the sampling distribution of
the
standard deviation of
example. the
known as the standard error
of the mean.
mean x
X+X, t . . . .

Where vx) v
=

f....
(x,)
n
n
2
no
t...... t
+
n
2
n n
The S.E. of the mean is
 The standard errors of the some of the well known
statis.
for large samples are
given below, where is the
n
sample size, o
the population variance and P is the population proportion and
Q 1-P. ni and n2 represent the sizes of two independent random
samplesrespectively.
SILNo Statistie Standard Error
Sample mean x

2. Observed sample proportion pP

PO
n
3. Difference between of two
samplesmeans (x 1- x2) n2
4 Difference of two sannple
proportions pi - P2 PQ + P:Q
n2

Uses of standard error

i) Standard error plays a very
important role in the large |.
sample theory and forms the basis of the testing of
hypothesis.
ii) The magnitude of the S.E gives an index of the precision o
the estimate of the parameter.
iii) The reciprocal of the S.E is taken as the measure
o
reliability or precision of the sample.
iv) S.E enables us to determine the probable limits within
which the population parameter may be expected to lie.
Remark:
S.E of a statistic may be reduced by
increasing the sampl
size but this results in corresponding inerease in cost. abour and
time etc.

4.4 Null Hypothesis and Alternative Hypothesis

Hypothesis testing begins with an assumption called*|
Hypothesis. that we make about a population parameter. S i
hypothesis is a supposition made as a basis for reasoning. T 8
conventional approach to hypothesis testing is not to construct
 6
about the population parameter but rather t. set

ngle hypothes is
p Two different hypothesis. So that of one hypothesis is aceepted
the other is rejected and vice versa

Null Hypothesis:
hy pothesis ofno difierence is called nuil hypethesis ard
usuaiy denoted by Hll hypothesis is the hyprothesis whieh
sted for
s te possible rjection
under the that it is assumption trie
Prof R.A. Fisher It is very usefui tool in test of sigificance
for example If we want to find out whether the special classes (for

Hr Sec. Students) after schoo! hours has benefited theclasses

studerts
after
tr

not We shall set up a nu!l hypothesis that " special

sc hool hours has not benefited the students
Alte rnativsHy pothesis:
the null
Anghypothe sis. which is complementary t
e s i s . s called an alternative hypothesis. usually denoted by
if want to test the null hypothesis that the
tor example. we

population has specified mean u, (say).

a

ie. Step 1: nuli hypothesisi ln:u-H

then 2. Aiternative hypothesis may be
1) H u, (ic u ju, or )
i1) H:uH
ii) H
the alternative hypothesis in (i) 1s known as a two- tailed
alternative and the alternative in (i1) is known as right-1ailed (1i) is

The eitings f
Anown as leti 1ailed alternative respectively.
enables to decide
ilernative hypothesis is very important since it us

left) two taiicd

nether we have to use a single tailed (right or or

test

.5 Leve! of significance and Critical value:

evel of significance:
probability with
esting a given hypothes1s. the
maximum

risk is called levcl

e would be willing t take
1cance ofthe test. This probability often denoted by "
eTer y specified before samples are drawn.
 J P e I amd T9pe-a &
Dbviously, the first two lead
ossibilities to errors.

riment,
In a statistical hypothesis testing experiment, a Type | a 1Ype error
error
committed by rejecting the hypothesis
hypothesis when
null
null
when it is true. On
On the
other aT ne
Oner hand, a Type II error is comm ted by not rejectin
rejectin,
(accepting) the null hypothesis when it is falsc
If we write,
aP(Type I error)
P (rejecting Ho| Ho is true)
ll error) P (Not rejecting Ho | Ho false) iS
Y p e =

n
practice, type I error amounts to rejecting a lot wnen it is gOod|
and type Il error may be regarded as accepting the lot when it is
bad. Thus we find ourselves in the situation which is described in
the following table.
Accept Ho Reject Ho
Ho is true Correct decision TypeI Error
Hois false Type ll error Correct decision
5.1 Large samples (n> 30): 8
The tests of significance used for problems of large
different from those used in case of small samples as
sampk
are
th
assumptions used in both cases are different. The followins
assumptions are made for problems dealing with large samples
(i) Almost all the
sampling distributions follow noma
asymptotically.
(11) The sample values close to t
are approximately close
population values.
The following tests are discussed in large sample tests.
) Test of
(ii) Test
significance
of
for
proportion
significance for difference between
proportions
 (ii) Test of significance for mean
two means.
(iv) Test of significance for difference between

5.2 Test of Significance for Proportion:

Test Procedure
Set up the null and alternative hypotheses
Ho: P=Po
H P # Po (P>Po or P <Po)
Level of significance:
Let a = 0 .05 or 0.01

Calculation of statistic:
Under Ho the test statistic is

Z=P-P

PQ
n
Expected value:

P-P N (0. 1)
PQ
N n
=
1.96 for a =
0.05 (1.645)
=
2.58 for a =
0.01 (2.33)
Inference:
(i) If the computed value of Zo s Zwe
accept the null
hypothesis and conclude that the sample is drawn from the
population with proportion of success Po
(I) f Z> Z we reject the null hypothesis and conclude that
the sample has not been taken from the
population whose
population proportion of success is Po.
PxampleIn 1
a random of 400 persons from a large
sample
4are females.Can it be said that males and females population
3 are in the
in the
population? Use 1% level of significance
Solution:
We are given
n 400 and
sample =
120
x =
No. of female in the
120
the sample =

P observed proportion of females in 400 03

Null hypothesis: are in the ratio to

The males and females in the population 53
3
i.e., Ho: P Proportion of females in the population
8
0.37
Alternative Hypothesis: G - 0.625
H:P*0.375 (two-tailed) 8
Level of significance:
a=| % or 0.01
Calculation of statistic:
Under Ho., the test statistic is

P-P

PQ

0.300-0.375
0.375 x0.625
400
0.075 0.075 3.125
V0.000586 0.024
Expected value:

Z=P-P NO.1) = 2.58

PO
Inference
Since the
calculated Zo> Z, we reject our null hypothesis at
and conclud that the males and females in
1% level ofsigniticance
a r e noI in the ratio 5:3
the population
Pample 2:
In a sample parts nmanulactured by a factory, the
ot 400

defective parts was found to be 30. The company.

number of 5o of their product is defective. Is the
claimed that only
however.
claim tenable?
ati:
Solution:
We are given

n=400
in the sample =
30
= No. of defectives in the sample
defectives
p propo.1on of

= 0.075
n 400

Null hypothesis:
P=0.05
The claim of the company is tenable Ho:
Alternative Hypothesis:
H P 0.05 (Right tailed Altenative)
Level ofsignificance: 5%
Calculation ofstatistic
Lnder Ho. the test statistic is Jale)
P
Z
PO

0 . 0 5 . 9 5 a uras\dek
0.075-0.050
means 4s/ non

400
0.025 2.27
0.0001187
 1A
Expected value:

P
Z N(O,1)
PQ
V n
= 1.645 (Single tailed) oru valu
9 f sqle taled
Inference:
Since the calculated Za
So
> Z. we reject our mull hypothesis
5% level of significance and we conclude that the company's claim

is not tenable.

5.3 Test of significance for difference between two proportion:

Test Procedure
Set up the null and alternative hypotheses:
Ho: Pi =P2 = P (say)
H:P P2 (P>P2or P <P;)
Level of significance:
Let a = 0.05 or 0.01

Calculation of statistic:
Under Ho the test statistic is

Z PP2 (P and P2 are known)

PQ
N n,

PP2 (P and P2 are not known)

where P ="P +n.p, X+X

n,+n n,+n,
Q-1-P
Expected v a l u e :
12
Pi-P N(0,1)
leS.E(p, -P2)
Inference:
) If Zo Z we
accept the null
hypothesis and conclude that
the difference between proportions are due
fluctuations.
to sampling
(ii) If Zo Z we reject the null hypothesis and conclude that
the difference between proportions cannot be due to
sampling
fluctuations

Example 3:
In a referendum submitted to the 'student body' at a
university, 850 men and 550 women voted. 530 of the men and 310
ofthe women voted 'yes. Does this indicate a significant
diterence of the opinion on the matter between men and women
students?
Solution:
We are given
n850 n2 550 X 30 X2310
530 310
0.62 P2 0.56
850 550
PtX 530+3100 = 0.60
n+n, 1400
0-40 (1-) -
(1-o.b
Null hy pothesis:
A: P P2 ie the data does not indicate
a significant
diflerence of

he
opinion
students.
and women
on the matter between
nmen

native Hypothesis:
. P#P2 (Two
lailed Alternative)
evel of significance
et a 005
 13

Calculation of statistic:

statistic is
test
Under Ho. the

PP:

Po n:

0.62-0.556

0.6x0.4 850 550

0.06 = 2.22
(ak
0.027
Expected value:

PP: N(0.1) = 1.96

P
Inference
Since Z, >Z we reject our null hypothesis at 5% level of

significance and say that the data indicate a significant difference of

the opinion on the matter between men and
women students
Example 4
In a certain city 125 men in a sample of
selfemployed. In another 500 are found to
city. the number of self
in a random sample of 1000. Does this employed are
greater population of self
indicate that there
first? employed in the second city than in
D
Solution:
We are given
n= 500
n=10000 N = 125
= 375
 175
125 0.25 P2
T000
0.175 14
$00
125+ 375

500 1000
500

I500 3

Null bypothesis:
:P P There is no signilicant difference betwcen the two
population proportions
Alternative Hypothesis:
PP(left tailed Alternativc)
Levelofsignificanee: let a- 0.05
Caleulation of statistic:
nder Ho. the test statistie is

PP

0.25 0. 375 125

A8
026

3500 1000

Fxpected value:

N0,) o5
 hypothesis
at 5% level.of
Inference: null
> Ze we reject the difference betweeen
Since Zo significant
there is a

significance
and say that
proportions
the two population
Example 5: to 200 people. On the
examination
was given
A civil service the upper 30% and
divided into
basis of their total
scores, they were
40 of the upper groun
70%. On a certain question On the basis of this
the remaining
answered correctly.
and 80 of the lower group for discriminating the
likely to be useful
question, is this question
tested?
ability of the type being
Solution:
We are given
30x 200 70x 200 =
140
ni 60 100
100
X2 8 0
X 40
40 80
-
P2
PI 3 140 7
40+80
n, +n 60+140
120 6
200 10

=1-P=1
6 10
Null hypothesis:
Ho: P- P; (say) The particular the
abilities of two groups. question does not discriminate
Alternative Hypothesis:
H: P P2 (two tailed
Level of significance: Alternative)
Let a = 0.05

Calculation of statistics
Under Ho, the test statistic is
S.4Test of significance for mean: 6
Let X, (i 1,2...n) be a random sample of size n from a
POpulation with variance o, then the sample mean x is given by

-(X1tX2t....Xn)
E(x)=
V(x) =
V[-(x1t x2 t..Xn)
 17
I(V(xi)+V(x)+...V(xn)
na2=
no
n

S.E(X)=-Vn

Test Procedure:
Null and Alternative Hypotheses:
Ho:u =Ho
Level of significance:
Let a = 0.05 or 0.01

Calculation of statistie:
Under Ho. the test statistic is
x-E(x) =

S.E(x) o/n
Expected value:

N0,1)
G/Vn
=
1.96 for a =
0.05 (1.645)
or

2.58 for a= 0.01 (2.33)

Inference
If Zo Z we
accept our null hypothesis and conclude that
the sample is drawn from
If Z> Ze
a
population with mean H= Ho
we
reject our Ho and conclude that the
not drawn from a
population with mean
sample is

j =

Ho
Example 6:
The lifetime of 100 fluorescent
mean
by a
company is computed to be 1570 light bulbs produced
deviation of 120 hours. hours with a standard
If u is the mean
produced by the lifetime of all the bulbs
company, test the
hypothesis u=1600 hours against
132
 1
the alternative hypothesis 1600 hours using a 5% level of
significance.
Solution:
We are given
x = 1570 hrs
1600hrs s=120 hrs n=100

Null hypothesis:
Ho: 1600.ie There is no significant difference between the
sample mean and population mean.
Alternative Hypothesis:
Hi: H 1600 (two tailed Alternative)
Level of significance:
Let a = 0.05

Calculation of statistics
Under Ho, the test statistic is

Sn

1570-1600
120
V100
30x 10

120
2.5
Expected value:

Z = N(O,1)

= 1.96 for a =0.05

5% level of
Inference reject our null hypothesis
at
between the
Since Zo> Ze we difterence

that there is significant

g i i c a n c e and say mean.
the population
anple mnean and
 19 whose
Example 7: introduce a
new car mnear
decided to
A car company
lower than that ot the existine
is claimed to be
tested tor petroi
petrol consumption were
taken and
n e w cars
A sample of 50 c o n s u m p t i o n tor the 50
car. found that m e a n petrol

consumptuon. It deviation of 3.> km per

was
standard
litre with a
w a s 30 km per s claim
cars
whether the company
litre. Test at 5% level of significance on the
is28 km per litre
consumption
that the new car petrol
average is acceptable.

Solution:
=28; n-50; s=3.5
We are given x = 30;
Null hypothesis:
28. i.e The company's claim that the petrol consumption of
Ho:
new car is 28km per litre on the average is acceptable.
Alternative Hypothesis:
H: 28 (Left tailed Alternative)
Level of significance:
Let a = 0.05

Calculation of statistic:
Under Ho the test statistics iss

s/n

30-28
3.5
50
2x V50
3.5
4.04
Expected value:

N(O,1) at a = 0.005
= 1.645
 Inference: 20 V1eluc.
Since the caleulated Zo
Ze we reject fhe nuil hypothesis
>
at
5% level of significance and conclude
that the company's claim is
not acceptable.
5.5 Test of
significance for difference between two means:
Test procedure
Set up the null and
alternative hypothesis
Level of significance:
Let a%
Calculation of statistic:
Under Ho the test statistic is

X2

If o =

o: =
o* (ie) If the
samples have been drawn from
the population with common S.D o then under Ho : Hi =

42

XI X2
Zo=

Expected value:
XI-N2 -N(0,1)
Inference:
S.E(X-Xx:)
)
IfZ s Z, we accept the Ho (ii) If Z,> Z, we reject the lHo
Example 8:
test of the breaking strengths of two different types of cables
Conducted using samples of ni = n2 = 100 pieces of each type
of cable.
 21
Cable II
Cable
, 1905
,-1925
230
o 40
provide suflicient evidence to indicate
Do the data
of the two cables?
difference between the mean breaking strengths
Use 0.01 level of significance.
Solution:
We are given
=1925 X=1905 40 2 30

Null hypothesis
Ho:H= 42 .ie There is no significant difference between the
breaking strengths of the two cables.
H:4 H2 (Two tailed alternative)
Level of significance:
Let a = 0.01 or 1%

Calculation of statistie:
Under Ho the test statistic is

Zo XI-X2

n n
1925-1905
40 30 T
4
Expected value:
90 100|

N (0,1) 2.58
-
O,

n,
 Inference: 22
Since Zo Z we reject the Ho Hence the formulated null
ie there
hypothesisis wrong
breaking strengths of two cables.
is a significant difference in the

Example 9:
The means of two large samples of 1000 and 2000 items are
67.5 cms and 68.0cms respectively. Can the samples be regarded as

drawn
5%
from
level
the population
of signiticance.
with standard deviation 2.5 cms. Test at

Solution:
We are given
n = 1000; n> = 2000
67.5 cms: , =68.0 cms G= 2.5 cms
Null hypothesis
Ho: = 2 (i.e.,) the sanmple have been drawn from the same

population.
Alternative Hypothesis:
H: 2 (Two tailed alternative)
Level of significance:
a 5%
Calculation of statistic:
Under He the test statistic is

67.5-68.0

2. 000 2000
0.5x 20
2.53/5
5.1
Expected value: 23
XI X2
Z N(0.1) = 1.96
1
n n2
Inference:
Since Zo > Z we
and conclude that the reject the Ho at 5% level of significance
population. samples have not come from the
same