Business Research Metbodolo2y

- (:hapter- Hypothesis Testing

(M. Con1-- IV Semester)
Prof. Arvind Kumar
Dept. of Commerce
University of Lucknon'
What is Hypothesis Testing?

Hypothesis is usually considered as the principal instrun1ent in research. The

main goal in many
research studies is to check ,,.rhether the data collected support certain ~taten
~ents or predictions.
A statistical hypothesis is an assertion or eonjecture concerning one or more
populations. Test
of hypothesis is a process of ~sting of the significance r~garding the
para1neters of the
population on the basis of sample drawrrfrotn it. Thus, it is also termed
as ·'Test of
Sign ifica nce' .~
..
In short, hypothesis testing enables us to make probability statements
about ~flU:~at~on)
p~ete r,,_ Th! hypothesis may not be proved absolutely, bu} in practic
e it is accepted if 1t as
withst
.......
ood a critica! testing. •
.

Points to be considered while formulating Hypothesis

• Hypothesis should be clear and precise.
--
/ /.

• Hypothesis should be qtpable'of being tested.

• Hypothesis should state relationship between variables.
• Hypothesis should, be lirrrned in s·cope' and must Fe specif.ic. • '
• Hypothesis should be stated as far as possib. le in most simple tenns
'-.. '- •
- -
so that the same is
easily understandable by all concerned.
• Hypothesis sHould be a'm~na~le to' testing within reasonable time.
• Hypothesis must explain·empirical referen~.
a
-
Types of Hypothesis:

There are two types of hypothesis, i.e., Research Hypothesis and Statist I
ical Hypothesis
l. Research Hypothesis: A research hypothesis is a tentative solutio
n for the problem
being· investigated. 'It is the supposition that .motivates the researcher
to accomplish
future coufseofaction. •In research, the researcher detennines \vheth
er or not their
supposition can be supported through scientific investigation.
" .... -

2. Statistical Hypothesis: Statistical hypothesis is a statement about the

,., t

population which
we want to verify on the basis of sample take~ from populatio~. Statist
ical hypothesis
is stated in -such a way that they may b~ evaluated by apprgpriate~statislic
al techniques.
Types of Statistical Hypotheses
.
There are two typ~s of statistical hypothese~; r~L l lvJ
. ~ ) -"'~ 6N'T
s-·· .. ·,, s~ "'!'l.
---
between a parameter and a specific value, or-tha
parameters.
• ----
1. Null Hypothesis (Ho) - A statistical ..hy.pothes1s that statei
- t there is·no differe
•
l' .'1 \lLA-\,A-1
that there is no difference
nse between two
'

.. I . ~- •,
: , I

"'1 1ZJ-AAL
-----,= =~=~---
_ _ ___...._ •;:::;cu ~.::cz :e~--: '..'.:.- l
 \lt rr 11 1u ,t lhpothC"
'\h lll , or Ila ) \ ~t~tt,ll\31 h,t x,t hl"
ol 1 ,,,ncra-..; tx- tw ttn n l'-1 '" thot ~to k~ the ~\l~h.
r1'1" td l\00 'O'-'C
\l 'lX ' h'- , all~.
J, tlc ttn e< bc t\\ 'ttn tw "'r ~tll\."' thl t th~T\'
o f'3 "r " 'd~ IS a
\ "· ma tt, c h~ l''t
m('!lnm~ ot the: null h, f" hl "'" ,, l't'CDtN m o
nc ga tt, '-'
"~'''
,uN '-l5 C \\ ( "in t
,c, \{' \ th< h, P'' 'h" "''' that th~
f"'f'Ul~th.,n mc:rn (~•)
h,'{',th(-..,'<<l 1rn ;3n \ \h•
} • l 00 1 h~n ,, "\,
,s equul to the
JX'rul:it.,)n ml.ln ,, '-"l uh l ~) thnt the nu ll h) pothc-;1~ ,s th:H the
u.1\ to lhl' h) P'' th" "''l
'\l mt Jn 1()() and ~,m
bolicn\ly "c c.nn

l f ou r ~m p\ e results
Ho. µ µ uo -= 100 ~ \; 4 -f40
do not su pp on this nu
som~thing els r \S true. \\ hyrothesis. "c shou
\\~hat WC conclude rc_ ld conclude that
ahem3ttYc hypothesis iecting)thc nu\\ h)rpOthesis
. ln other words. the is k~own ll$
referred to as the nltem set of ohcmotives to the
ati\'e hypothesis. lf "c nu \\ hy pothesis is
accept H,\. then we arc
if we reject Ht\. then we
are accepting H,. For rc_ kcting H, and
possible alternative hy H,,: µ = µH" = 100. we mn
potheses as fo\lows: y co nsider three

(The alternative hypothesis

is that the population mean
, is not equal to 100 i.e
., it may be more or les
s than 100)

tToe alternative hypothe

sis is that the population
is greater than 100) mean
(The alternative hypothe
sis is that the population
is less than l 00) mean
Th e null hypothesis
an d the alternative hy
drawn tth e researcher pothesis are chosen be
must avoid the error fore the sample is
he/she collects and the of deriving hypotheses
n testing the hypothe from the <latn that
null hypothesis. the fol ses from the same data)
lowing considerations . In the choice of
1. Alternative hypothe are usually kept in view:
sis is usually the one wh
hypothesis is the on e ich one wishes to prove
which one wishes to and the null
represents the hypothe disprove. Thus, a null
sis we are trying to hypothesis
represents all oth er po reject, and alternative
ssibilities hy pothesis
2. Null hypotheses sh 1 ,
ould always be specif
about or approximate ic hypothesis i.e., it sho
ly a certain value. uld not state
3. ln testing hypothesis
, there are two possible
• Reject Ho and acc outcomes:
ept H, because of suffic
favour of H, ; ient evidence in the sum
p le in
• Do not reject Ho be
cause of insufficient ev
idence to support H 1.
BASIC CO NC EP TS
CONCER1'1NG TEST
ING OF HYPOTHES
1. Th e level of signif ES
icance: This is a very
testing. lt is always so im po rtant concept in the co
me percentage (usually ntext of hypothesis
care, thought an d rea 5%) which shou 1J be-
son. ln case we take the chosen with great
implies that Ho will be significance le\ el al 5
rejected when the sam pe r cent. the r this
pling result lie ., obser • ·, •• ,
a less than 0.05 proba Ye d eviden~e) has . : :
bility of occurring if Ho .
of sionificance means is tru e. In other ·words, th~ 5
that researcher 'is will per cent lev d • .- . .
0
rejecting the null hypo in~ to take as much as .. '
.. - thesis when it l H,,) ha -
ppens to be tru~. Thus,
a 5 per c:ent risk of
• • - ·'
the signi ftcunc~ • -
:
''

I ..
 level is the maximum value of the probability of rejecting Ho when it is true and is
usually detennined in advance before testing the hypothesis.
2. Decision rule or Test of Hypothesis: A decision rule is a procedure that the n:scarchcr
uses to decide \\1hethcr to accept or reject the null hypothesis. The decision rule is a
state1nent that tells under what circun1stances to reject the null hypothesis. The decision
rule is based on specific values of the test statistic (e.g .. reject Ho if'Calculated vnlu~ >
table value at the saine level of significance)
3. Types of Error: In the context of testing of hypotheses, there are basically two types
of errors we can make.
a. Type 1 error: To reject the null hypothesis when it is true is to make what is
known as a type I error. The level at which a result is declared significant is
known as the type I error rate, often denoted by a.
b. Type II error: If we do not reject the null hypothesis when in fact there is a
difference between the groups, we make what is known as a type II error. The
type II error rate is often denoted as p.

In a tabular form the said two errors can be prese_nted as follows:

I'. • ' - - •ffl -
IBl ccept Ho Reject Ho
Ho (True) • Correct Decision Type I error
(a error)
Ho (False) Type II error Correct decision
(P error) .

-
One- tailed and Two-tailed Tests: A test of statistical hypothesis, where the region of
rejection is on OJ!lY one side of the sampling distribution, is called a one tailed test. For
example, suppose the null hypothesis states that the mean is less than or equal ,to 10.
The alternative hypothesis would be that the meaii7s greatec than 10. The region of
rejection would consist of a range of numbers located on the right side of sampling
distribution i.e., a set of numbers greater than 10. •
A test of statistical hypothesis, where the region of rejection is on both sides of the
sampling distribution, is called a two-tailed test. For example, suppose the null
hypothesis states that the mean is equal to I0. The alternative hypothesis would be that
the mean is less than IO or greater than I0. The region of rejection would consist of a
range of numbers located on both sides of sampling distribution; i.e., the region of
rejection would consist partly of numbers that were less than IO and partly of numbers
that were greater than l 0.
 Rjght-tail test
H., I(, man~ probahle

Ha:µ> value
0

Left-tail test
. Hd is more pr obJblr

Ha: µ.. < value

0

H0 is more µ1 obable
Two-tail test

Ha: µ =I= value

0

Procedure of Hypothesis Testing

Procedure for hypothesis testing refers to all those steps that we undertake for making a choice
between the two actions i.e., rejection and acceptance of a null hypothesis. The various steps
involved in hypothesis testing are stated below:

1. Making a formal statement: The step consists in making a fonnal statement of the
null hypothesis (Ho) and also of the alternative hypothesis (Ha or H1). This means that
hypotheses should be clearly stated, considering the nature of the·research problem.
2. Selecting a significance level: The hypotheses are tested on a pre-detennined level of
significance and as such the same should be specified. Generally, in practice, either 5%
level or l % level is adopted for the purpose.
3. Deciding the distribution to use: After deciding the level of significance, the next step
in hypothesis testing is to detennine the appropriate sampling distribution. The choice
generally remains between nonnal distribution and the t-clistribution.
4. Selecting a random sample and computing an appropriate value: Another step is
to select a random sample(s) and compute an appropriate value from the sample data
concerning the test statistic utilizing the relevant distribution. In other words, draw a
sample to furnish empirical data.
s. Calculation of the probability: One has then to calculate the probability that the
sample result would diverge as widely as it has from expectations, if the nu11 hypothesis
were in fact true.
6. Comparing the probability and Decision making: Yet another step consists in
comparing the probability thus calculated with the specified value for a, the
significance level. If the calculated probability is equal to or smaller than the a value in
case of one-tailed test (and a /2 in case of two-tailed test), then reject the null hypothesis
(i.e., accept the alternative hypothesis), but if the calculated probability is greater, then
accept the null hypothesis.
The above stated general procedure for hypothesis testing can also be depicted in the fonn
of a cart flo\v-

l
Tests of H yp ot he se s h •on (tec hni•ca lJy de sc rib ed as
nu ll
• • e as~umpt1
· mi·nes the vahd1ty of t t th e va lu e of a
Hypothesis testing deter s~ be tw ee n tw o co nf licting hy po th es es ab ou et he r
to choo sis ~f ~-ample da ta, wh
hypothesis) with a vi ew lps to de cid e on the ba
pothesis testing he ve de ~e lo ~e d
population parameter. Hy ely to be tru e or false. Stat1st1c1ans ha
population is lik tes tin g
a hypothesis ab ou t the the te~ ts of sig nif ica nc e) fo r th e pu rp os e of
s (also kn ow n as
several tests of hypothese I
be classified as:
of hypotheses which can l
an d
ram etr ic tes ts or sta nd ard tests of hypotheses; se s ...
a) Pa bu tion-free tes t of hy po the
etr ic tes ts or dis tri
b) Non-param we dr aw
pr op ert ies of the pa ren t po pu lat io n fro m wh ich
assume certain ge ,
Parametric tests usual1y fro m a no rm al po pu lat io n, sa m pl e siz e is lar
e obseivatio.ils co me riance, etc ., mu st ho ld
go od be fo re
samples. Assumptions lik ram ete rs lik e me an , va no t
pulation pa ar ch er ca nn ot or do es
assumptions about the po ua tio ns wh en the re se
used. But there are sit sta tis tic al me th od s fo
r te sti ng
parametric tests ca n be ch sit ua tio ns we us e
want to make such assu
mptions. In su t de pe nd on an y
etr ic tes ts be cause su ch tes ts do no
lled no n- pa ram tri c tes ts
hypotheses which are ca pa ren t po pu lat ion . Be sid es , mo st no n- pa ra1 ne
assumption ab ou t the pa
ra1neters of the en t eq ui va le nt
ere as pa ram etr ic tests req uir e me as ur em
ordinal da ta, wh on s th an
assume only nominal or , no n- pa ram etr ic tes ts ne ed mo re ob se rv ati
to at least an inteival
scale. As result pe II errors.
tes ts to ac hie ve the sa me size of Ty pe I an d Ty
parametric
ET RI C TE ST S ts ar e ba se d
IM PO RT AN T PA RA M ; (2 ) t-t est ; an d (3 ) F-test. Al l th es e tes
te~ts ~re: ( l) z-t est ib ut ed .
The important ~arametric ur ce of da ta is co ns ide red to be no rm all y di str
ah ty 1.e., the so J·ud · th
on the assumption of no nn dis tri bution an d is us ed fo r e
1. z- test: It is ba sed on the no • I rm al pr ob ab ili ty
• lar ly the me an . Th is is giang m os t
• ~articu
s1g• ru"fi1cance of severa_1 statJstica mea~w-es, is test is us ed ev en wh b•
m om .a
J
rch stu die s. Th end' ·b ·
frequently used test m resea • appli.cable on the pr es um pti on th at su ch a 1s tn ut1on
'b • or t-d'istri'b uti•on 1s · .
d'1stn"b uti•on as , n , bec om es •lar ge r • -7-f es t 1s ge ne ra 11y• us ed
•
d1stn ut1on
• t a I .
tends t o approx1ma e normf • to so me hy po the sis ed me an fo r th e po pu 1an on in
• h an o a sa mp le .
1or comparing t e me
r. . . n ... t· t ·
I sa mp Ie, or w h en po pu lation varianc•.e 1s kn ow • ...d- es d1s a 1so us ed fo. r
f
case o arge• 'fi · een me an s nf two in ep en en t san1ples in
•d •
JU g1ng he s1gn1
1cance of difference betw ••
 cctse of large samples, .or when population vari
ance is known. z-test is also used for
n or for
~om~aring the sample prop ortion to a theoretical value of population proportio
pendent samples when n happens to
Judg,ng the difference in proportions of two inde
ing the significance of median, mode,
be large. Besides, this test may be used for judg
sures.
coefficient of correlation and several other mea
ed an appropriate test for judging
2• t- test: It is based on t-distribution and is considersignificance of difference b~tween
the
the significance of a sample mean or for judging
ple(s) when population variance is not
the means of two samples in case of small sam
sample as an estimate of the population
known (in which case we use variance of the
use paired t-test (or what is known as
variance). In case two samples are related, we
e mean of difference between the two
difference test) for judging the significance of-th
the significance of the coefficients of
r:lated samples. It can also he used for judging
s1mple and partial correlations.
to compare the variance of the two-
3• F-test: lt is based on F-distribution and is used context of analysis of variance
in the
independent samples. This test is also used.
e than two sample means at one and the
(ANOVA) for judging the significance of mor
significance of multiple correlation
same time. It is also used for judging the
coefficients.
Non parametric Tests
out ir
isn't nonnal. Therefore, the key is to figure
Non paran1ctric tests ar~ used when the <latu ss
non-para1netric test you are likely to c01ne acro
you hav~ nornm\\y distributed data. The only the
ever. there are scYeral others. For cxnmpk:
in clcn1entary stnts is the chi-squure test. How fv1ann
native to the One-way ANOV A and thr
Kruska\ Willis test is the non-paran1ctric riltcr
th~ two-san1pk t t~~t.
Whitn~y is the non- para1netri~ alt~rnatiY~ to
lll11stration 1:
it be
to have a mean height 67.47 inches. Can
A sample of 400 male students is found and
popula~jon with mean height 67.39 inches
reasonably" regarded as a sample from a large
level of significance.
standard deviation 1.30 inches? Test at 5%
67 .39
the mean height of the population is equal to
. Solution: Taking the null hypothesis that
inches, we can write:
Ho: µHO = 67.39"
H,: y HO=/= 67.39''
1
= 1.30" , n. 400. Assuming the populatio
n to
and the given information as X = 67.47~ , _<JP
z as under:
be nonnal, we can work out the test statistic
Z-T CST

. --
J - • ,

·- r~·1•. 111 c:' • ., .. 1 1-' 1 •

,I ')( ;1f I IJ,. " I II f I
, , f I •• •
•-'flrn

! I , ·V • 1; JI ( •: : 1 I I,I ' C '. • I

. .... t , I ••I t,I •( t , I • I ·, • '

.... 11
 - -
Hence, z .::: 67 •47 ~6 .39 _ o.oa
_
. . . 1.30!'140o - 0.065 = 1.23 I
As Ha 1s two-sided in the given qu • t· . . . . ,. ••
• • . .
es ion we shall 6e app ly1ng . _J • •
the reJect1on regions at 5% level 0 f . a two-ta1leo test for detenn1n1ng
area table: signi~cancc which.comes tO as under. usin,g no~al curve

R: I z I> 1.96
The observed value of z is I 231 whi h . . • • •
HO is accepted. We ma co~ c ts m t~e acceptance regi_on-~Ln~J z I > 1.96 and thus~
d <l'.t h b Y elude that the given sample (with mean height =·67.47") can be
rega~ e • ave een. ta~e:o fro,n a P.opulation· with mean height 67.39" and standard deviation
I.30 at 51/o level of ~1gn1ficance.
Illustration 2.

Suppose \re ~rt mtcr~lr d 10 a popuJation of 20 1udnst1 ,a) unn~ of the ;ame s1Ze. all of \,·tuch are
expeneu_clll~ ex5ts~~·e labo1tr turnover proble111.s. ni~ past r~Qrds _s_hoi.,· ~-t the :'.llean o~ the
I
4 • .............
•

disttibullo~ of ai~u~, fllruo\·er 1s 3~0 en1ployee~. \\·itll a standfu-n dt"Vi;ltton. l)f 75 einplov~e..,_ l
s;unple of) of tbe,Je 111du~trial tuuts is taken at random \\"!uch eives a mean of annual n1rno\·e-r a.s...300
e1uplo)·ees. Is the sample mean ronsi~r~iu \V~ t~1e populatio...n mean? Te~t at ~ 'o !c,.-cl- ---
0

\ ' i ·1 Taking the null hypotbesi~.that ~be populaJ1on mean~ 320 employee':;_ \,;e- t:au \1.T1Te:
)}J

H,, :. ir-;i: =' 310 eniployecs

ff3 : ~t H 3.:0 e1nployees
~11d the 2i\·en ulfonnat1011 as under·
·.r = 300 eruployee5. cr;, = 75 employee~

..~ssununp. the population to i,e nomuL \l.'e cm \\'Ork out the test ~tatistic :: 3S under

.
300 _· 310
__ ____ - -
_..;.. - ~o _..,

I 33.54) 1.888 I
• 75, ~/5 _,< ~'I ~Q - 5,. l,,~I 20 1i
. ,

= - 0.67
• - "•
~ .~s H 1; m·o-stded g1,·en question. \re <;hall apply J h\'O-tuled rest for deternll1llng the
w tl1~
retection ;e~ious at 5v O le\ e! of significance \1:lucl1 co111es to JS under. ustng no~u: c~tr\·e are.1 table-
• I..
1

• R _ 196
11lr obserYed ra}uc qf: 1~ --0 67 ,\'hich 1~ 111 tht;_ ~cceptance reg_ion since R : 1~:6 anj thn;_
H, 15 :i~;:epted :ud -.re may conclude lliJi tbc ,;uuplc mean b cons1qem WHh p0pulc111oc mean I e
the populallon mean 3~o IS t:pported bv S,lntple results
 -
111ustradon: 3 ,
.
RaJU Re sta ~a nt ne ar the raihvay
.. . '
station at F alna has ~ m having av
day Because of th e ~v el op m ~t ~a ~c s~l~s of 50 0 tea cu ps
of bus Sb nd nearby. it ,x pt ct s to inc
1 ctay,;, af ter the.- s_ran
o f~ bu s stand. the.- daily sale.-<, we re~se its sales. Du rin g th e f~ t
re.- .a!; undc.-t: . , '"I
5? 0. 57 0. 49 0 615.5?5.SS0.570.46
0.000.580.530.5~6 ,.
On.the \);m~ 0 ~ ttw.
-li5-t s per cent le\re\ or st ~f \IU ~~ O A. ca n one conelu ~ that Ra
ic ar lc c. ju Rp tau r.; nt ' sales_have ip cr en cd
'
1
"-f1!r1U(' • Ta ki ng th e ~b y1 >9 th es
incre:1<,e:d un le ss provcll,'Wc.-'can 1s that ~ales a\·tta ge 500 tea· cups P..'1 <Liy 3? d tM y lu ;·e no
"\\"rite: ••• ' ' t
·• - ,.._ _
H0 : µ =- ~p s" ~·da _y
H • ~1 500 (ac; \\,e \\·ant to co nc ~,
ol
lu ~ tha t sa l~ have in cr c~ ed )
.:\5 the 5,ample size 1s small and
the population standard deviation is
assuming no m la l populat1on an d not kno~"U. \Ve shall us e r-t eSt
shall \\·otk out tbe te.;t stntist1c r a\
:

rro find' ."\ , ' cfl

-
\ 'v \a
' nd \\"r t Ill•.He the'foUo\v;no
iw.- t, computations:) ---
·- --- - - - -
---- - ) ' ,~-=--
s.No.
i;l
--- - r, ' '

.it; X· - XS) - l-~i-_

.
'
~-~~_:_-:--:--_.::__(:_ I
~.., ~~-~~:..___-~,7 ii - ~ ~
,
1 '
sso
') -
' .&
3 ' ~'
490 -~
 .. .-
table of r-d1~tnbution for 11 dtgrees of freedonr
R r 1 i96
reg1011 and rhus Hr, 1s re3ecred at 5
reJ~cuon per
Thr OPJtrrtrJ 1;J )11 ttJ/11r_ J.>)J'<rfttc!i rs ri:1 (fie
,,,e can con clu de f the }an1plt cLira indicate rha t_&JV rr,TJlllaIJf ·§
cent le\'el of s1giufican ce and thtl ,.

salei lure 1ncrea~ed.

I
Illustration 4
sho ps in to\, ns are take~ for a ne\ r pro duc r ,r1th tbe follo',rlllg re)ulrs
Sai11ple of ~ales lll sunilar t\\'O

- • , Tol4'11 ~\lean ;ale.s

57
I'ariance
5.3
Si:e, of;an1ple
-
)

4.8 7
61
nce 111 s3Je s in the nro ro,r ns? t 1
se 5 per cenr Jerei of ~1gn1ficance
Is there any rridence of differe
r,ro ~an1ple-s .
for testing rhis difference benreen the 111eans of
l h~p oth esis the n1e am of nro pop pl2t 101 1s do not_diffrr v,·e can ,rnte.
, , • Taking rhe nul rha t
H0 : µ! =~' ,. f • • • ,r, • • . .. :

md tlle giren infom1ation as follo"·s •

Sm,,:oJe fro ", to1t11 .A

as sa,nple one X11 = 57 a:...., = :,_-3• nl =)

SanlJJle frorn f(!M11 B

As sa,npf! rwo x" = 61
·-
- e o f.~lp le~ t·
rlle- siz
Swee ~n .the given que stion var iances of the population are not kno,,·n .,"d
u"
., • '
1, . c •0
•. •

11 b 11 use r-rest 1or differe nce

•.L' -
m rne an:. assunun,,. the ~ be and can
sn1.1 . \,·e s .a =----popu ~Jiou~ i"O rmJ l
ivork out the test statistic /asunder·
- "'
_
-
- ' r= r-
1 1 1
\'~ - n~
with d.f = (n.. - n.- - 2J