Professional Documents
Culture Documents
HYPOTHESIS ???
is formally stated expectation about how a behaviour operates. is a proposition that a researcher wants to verify. is a conjectural statement of the relation between two or more variables.
P - Value .????
Probability Value or p - value is the probability of observing a sample outcome even more extreme than the observed value when the null hypothesis is true. The smaller the p - value, the smaller are the chances that variations are caused by chance/random factors. It is also called observed level of significance. It provides an alternative way to decide whether a null hypothesis is to be accepted. It has following advantages and thats the reason mostly statistical softwares are giving printouts with p - values: it allows a decision maker to use his/her own level of significance and make decision accordingly once sample results are available with necessary statistic it provides very precise information about the highest level of significance at which the null hypothesis must be accepted.
preparing ourselves for the necessary backdrop to take forward a solid move for
First, we take
HYPOTHESIS
2 2 2
2 2 2
2 2 2
2 2 2
Count
2 2 2
2 2 2 YS E N O
Can we conclude from the above data that there is a significant difference between those who say YES and those who say NO?
Before starting or after the journey were you or will you stay in hotel / Dharmshala etc.
YES NO
B i n o m i a l Te s t C a te g o ry N D o y o u t h in k t h is G rvoe un pt 1 e YES 2 68 ( C o m m o n w e a l t h G a m e s) i s a n o p p o r tu n i ty Gt or o u p 2 2 32 sh o w c a se r i c h I n d i a n N O c u l t u r a l h e r i t a g e Tt oo ta o r l d ? wl 5 00 Y e s/ N o a .B a se d o n Z A p p r o x i m a t i o n . O b se r v e d A sy m p . S i g . P r o p . T e st P r o p( 2 -t a i l e d ) . .5 4 .4 6 1 .0 0 .5 0 .1 1 8
a
a. b. c. d. e.
Athletics Cricket
Count
8 0
6 0
4 0
Hockey WWF
2 0
0 A le th tics C e rick t H ck y o e W F W O e th rs
S o y u lik m s to w tc o T p rts o e ot a h n V
C se we h db F E a s ig te y R Q
Can we conclude for the above that people watch different types of sports equally?
a. 0 cells (.0%) have expected frequencies less than 5. The minimum expected cell frequency is 52.0.
Is there any statistically significant difference in the state of ENO before and after joining a health club?
Is there any statistically significant difference in the state of the person before and after taking coffee?
Related Samples...
. are defined as those where the observations in one has some relation or influence on those of the other sample.
Research Project:
What to do ? IMPACT OF PLAY SCHOOL EDUCATION ON THE PERSONALITY OF CHILDREN CHILDREN
One of the research issue in it whether children joining play school become more independent and confident
+ C A
+ B D
A AND D SHOW CHANGES BETWEEN RESPONSES. IF THE TREATMENT HAS NO IMPACT, THEN HALF OF (A+D) MUST CHANGE IN ONE DIRECTION WHILE OTHER HALF SHOULD
(CONTINUED)
NULL HYPOTHESIS IS - WHETHER THE PROBABILITY OF A, P(A), IS EQUAL TO PROBABILITY OF D, P(D); i.e. P(A) = P(D) = 1/2. FOR IT, WE HAVE 2 = (A-D)2/(A+D) WITH df = 1 WITH THE CORRECTION FOR CONTINUITY2 = ( A-D 2/(A+D) WITH df = 1 -1) IF THE EXPECTED FREQUENCY IS SMALLER
Were you satisfied with the billed amounts prior to privatization? YES NO Were you satisfied with the billed amounts after privatization? What is the thing YES NO
YES/NO
110
100
YES/NO
100
90
90
80
80
70
Count
70
70 YES NO
Count
75
60 Y ES NO
B efore the present m ess contractor, w the food as oily and/or S picy? 1 2
b
60 45
T est S tatistics
B efore the present m ess contractor, w the food oily and/or as Spicy? & Is the food of the present m ess contractor oily and/or S picy? N C hi-S quare Asym S p. ig.
a
Group I Group II
2the exact
Group xGroup y A B C D
probability of observing a
A C D + B + A B p = N + A B
On this basis and referring to necessary Table, one can decide whether H0 is to be accepted or rejected.
From the following data, can we conclude whether a particular Fund is performing better than the other? PERFORMING PERFORMING
BETTER THAN MARKET 25 22 WORSE THAN MARKET 15 25 SECTOR FUNDS BALANCED FUNDS
Total 2 2 2 2 2 2
FUNDS Total
P earson C hi-S quare C ontinuity C orrection Likelihood R atio Fisher's E act Test x Linear-by-Linear A ssociation Nof V alid C ases
a. C puted only for a x table 2 om 2 b. 2 cells (. % have ex 2) pected count less than2 The m . inim ex um pected count is 22 . 22 .
DECISION ?????
Research Project:
What to do ? QUALITY OF MANAGEMENT IN PUBLIC INSTITUTIONS INSTITUTIONS
One of the research issue in it Do public have different experiences in dealing with public institutions like DDA, MCD, NDMC, etc.?
NOMINAL DATA
NOMINAL DATA
WHERE
K = NUMBER OF COLUMNS; Ri = TOTAL OF ith ROW; Cj = TOTAL OF jth COLUMN; S C = SUM OF TOTAL SCORES; = MEAN OF COLUMNS TOTAL.
Assume that five members A, B, C, D, and E of a mountaineering club each attempt three different rock climb at each of which they either succeed or fail. The outcomes are shown 0 as success. MEMBERS below readB as fail and 1 D A C E
CLIMB#1 CLIMB#2 CLIMB#3 1 1 0 1 0 1 0 0 1 0 1 1 1 0 1
T s Saisi s e t t t tc N Cc r n Q o ha 's d f Ay p S . s m. ig 2 22 a 2 .2 2 .2 2 2
a 2 tr ae a as c e s . is e t d s u c s .
DECISION ?????
Research Project:
What to do ? IMPACT OF Sarve Shiksha Abhiyan ON THE RURAL DEVELOPMENT IN INDIA INDIA
One of the research issue in it- Whether no. of children from different castes in villages going to primary schools, middle schools and high schools are different.
NOMINAL DATA
MORE THAN TWO SAMPLES (UNRELATED) THE CHI - SQUARE TEST It is an extension of 2 independent samples. The null hypothesis is whether there exists a significant difference between the K independent groups.
Assume that you want to judge the Financial Analysts ability to predict correctly share prices in the market. For that you collected the following data for 100 days aboutthe prediction of 5 analysts about a particular FORECAST share. WITHIN ACCEPTABLE BEYOND ACCEPTABLE
RANGE ANALYST S A B C D E 35 45 36 48 50 RANGE 65 55 64 52 50
FIN N L A CIA A A S N LY TS
A B C D E
Total
Total 22 2 22 2 22 2 22 2 22 2 22 2
C i- q a eT ss h Sur e t V lu a e 22 2 .2 22 2 .2 22 2 .2 2 2 2 d f
a
Ay p S . s m. ig (2 e ) - id d s 2 2 2 .2 2 2 .2 2 2 .2 2 2
DECISION ?????
BINOMIAL TEST
2 CLASSES
McNemar Test
HYPOTHESISTESTING RELATED TO
Research Project:
What to do ? PUBLIC SCHOOLS IN DELHI A STUDY STUDY
One of the research issue in it Do public have different preferences in sending their wards to private public school, government-aided schools and government schools?
2 2 2 2 2 2
2 2
2 2
2 2
2 2
2 2
Count
For One Tailed Test D = MAXIMUM Sn1(X) - Sn(X) where, Sn1 = Proportion of Cum. Frequency of distribution one; Sn = Proportion of Cum. Frequency of theoretical distribution;
What according to you should be the ideal period of review? a) 1 Month b) 2 Months c) 3 Months d) 4 Months
a,b
Kolmogorov-Smirnov Z Asymp. Sig. ( 2 -tailed) a. Test distribution is Uniform. b. Calculated from data.
DECISION ?????
Research Project:
What to do ? DETERMINANTS OF MOTIVATIONAL
LEVEL OF THE FLOOR-LEVEL WORKERS A CASE STUDY OF PUNCHKULA PLANT
One of the research issue in it Is the motivation level of floor works at PUNCHKULA PLANT higher than that of the industry?
Celebrity increases appeal of a product 1. Strongly Disagree 2. Disagree 3. Neither Disagree nor Agree 4. Agree 5. Strongly Agree What is the thing
CL BI Y E E RT I CESS NRAE P OUT R DC APA PEL N Md n e ia C i-S u re h qa d f Ay p S . s m . ig Y te ' C n u a s o tin ity C rre tio o c n 2 2 2 22 2 . 22 22 .2 2 .2 2 2 22 22 .2 2 .2 2 2 2 2 2 2 2
C i-S u re h qa d f Ay p S . s m . ig
a G u in V ria le V R . ro p g a b : A
DECISION ?????
Research Project:
What to do ? A STUDY OF EFFECTIVENESS OF
TRAINING PROGRAMMES CONDUCTED BY NSE FOR SHARE BROKERS AND SUB-BROKERS IN INDIA
One of the research issue in it Is the comfort level of brokers/sub-brokers in trading mechanism increased after training?
ORDINAL DATA
THE SIGN TEST
TWO SAMPLES(RELATED)
It is applicable for two samples which are related. It tests whether there exists any difference between the observations of two related samples. NULL HYPOTHESIS : P(A < B) = P(A>B) = 1/2; i.e. if there is no difference in the related observations then number of changes of higher values over other must be equal to number of changes of lower values over other. For Small Sample: USE BINOMIAL DISTRIBUTION AND for large samples: USE NORMAL DISTRIBUTION with MEAN = (1/2)N AND VARIANCE = 1/4 N where N = total number of signs.
ORDINAL DATA
THE SIGN TEST
TWO SAMPLES(RELATED)(continued)
To apply the Sign Test, our data should be presented as follows and work out the sign of difference among the values of Sample A and Sample B.
Sample A Sample B 1 2 3 : : Sign
For small sample, USE t STATISTICS and for large samples use NORMAL
PROBABILITY DISTRIBUTION with MEAN = ( N ( N+1 )/4 ) AND VARIANCE = (N(N+1) (2N+1)/24) where N = Total Numbers of Pairs less dropped outs.
Do you and your boss go the joint goal setting meeting after doing proper groundwork based on MOU targets for the department? (Tick one)
YOU YOUR BOSS
a) b) c) d) e)
N 2 2 2 2 2 2 2 2 2 22 2 2
a Y U B S <Y U . OR OS O b Y U B S >Y U . OR OS O c Y U B S =Y U . OR OS O
T t tt t s e Si i s a c s
a
YR S O BS U O - O Y U Z Am i .(- ie s p g t ld y .S a ) a . S Tt i ne g s 2 - 2 . 2 2 .2 2 2
DECISION ?????
a. Y U B S < Y U O R O S O b. Y U B S > Y U O R O S O c. Y U B S = Y U O R O S O
Tt tt t s e Si i s a c s
YR S OBS UO -O Y U Z A p i . ti d s . g a ) y S (- l m e a . b . 2
- 2 . 2 a 2 .2 2 2
B dne t e n. a on a r k s e gi a s v W oS eR se i oni n a Tt l x gd n c k s
DECISION ?????
Research Project:
What to do ? QUALITY OF MANAGEMENT IN INDIAN CORPORATE SECTOR
One of the research issue in it Is there any significant difference in the opinions of shareholders and the management about the quality of governance in India?
Another Problem
What to do ?
Research Project:
PHYSICAL FITNESS AMONG STUDENTS OF INDIA
with df = 1
Mn h eU a - in nW y t Wx W ic o lon Z AmS ( -a d s p i . t ie y . g l ) a .
MT Y OH NL U G IL S EL A / B I R Ns . 22 2 2 2 2 2 2 2 . 22 2 2 2 2 2 2 2 . -2 2 .2 2 .2 2 2
G pg a b:M I E? r un Vi l oi r e O .. a B . L
DECISION ?????
P Y IC L HS A F NS IT E S S OE C R 2 2 2 .2 22 2 2 2 .2 22 -. 2 2 2 .2 2 2
a G u in V ria le G O P GV R B E . ro p g a b : R U IN A IA L
DECISION ?????
There is no difference between median spending. Do you believe that there is no difference between the spending of Pre- and Post-Mobile Users?
USAGE DISTRIBUTION BEHAVIOUR AMONG PRE- AND POST-PAID MOBILE USERS
2 2
2 2
2 2
2 2
Count
POST-PAID
MONTHLY USAGE/BILL
There is no difference between median spending. Do you believe that there is no difference between the spending of Pre- and Post-Mobile Users?
P O S T O R P R E -P A ID ? * M O N T H L Y U S A G E / % w ith in P O S T O R P R E -P A ID ?
It tests whether the frequency distribution of two samples is same. An extension of one sample goodness of fit test.
N 2 2 2 2 2 2 2 2 2
DECISION ?????
Research Project:
What to do ? ACCEPTABILITY OF FIFTH PAY
COMMISSION REPORT AMONG THE GOVERNMENT EMPLOYEES
One of the research issue in it DOES THE REPORT HAS SAME DEGREE OF ACCEPTABILITY ACROSS VARIOUS CATEGORIES OF EMPLOYEES?
ORDINAL DATA
THE MEDIAN TEST
ORDINAL DATA
MORE THAN TWO SAMPLES (UNRELATED)
THE KRUSKAL WALLIS TEST- ONE WAY ANALYSIS OF VARIANCE
It is a test that is very useful in determining whether K independent samples are coming from the same population; that is to say, it tests basically whether the differences among samples signify genuine population differences. In it, first all observations must be replaced by ranks to be allocated to each observation on the basis of combined observations from all the samples. If Null Hypothesis is to be true then the sum of ranks for each sample should be significantly different. Then, the following test statistics is calculated-
ORDINAL DATA
MORE THAN TWO SAMPLES (UNRELATED)
(continued)
Rj2
where, K = number of samples; nj = size of sample j ; N = total number of observations in all samples; and Rj = summation of ranks in the jth sample.
The example taken is based on experimental designed 4 different groups of students have been taught differently by using 4 different techniques of teaching. Their test records are noted which are given below:
1 65 87 73 79 2 75 69 83 81 3 59 78 67 62 4 94 89 80 88
Kruskal-Wallis Test
Ranks TEST SCORES GROU OF STU P DEN TS GROU # 2 P GROU # 2 P GROU # 2 P GROU # 2 P Total
a , b
N 2 2 2 2 2 2
M ean Rank 22 . 2 22 . 2 22 . 2 22 22 .
Ts S tsi s e t t i tc a
TS ET SOE C RS 22 2 . 2 2 .22 2
DECISION ?????
Mr. Jayant Saxena is doing a research project on the academic excellence among Indian MBA students. For that, he has divided all the students into 3 categories Engineers and Science Graduates, Commerce and Economic Graduates; and others. He collected their final grade points that are out of a total of 5 points.
N 2 2 2 1 2 3 6 6
Ma Rn en ak 4. 4 58 2. 9 96 2. 7 51
Ts S ts c et t i t s a i
C- qa hSur i e d f A m S. s p i y . g a . b .
GAE RD P I T OT O SU N O 5 OT F PI S N 1. 5 4 5 2 2 .0 01
ORDINAL DATA
MORE THAN TWO SAMPLES (RELATED)
THE FRIEDMAN TEST - TWO ANALYSIS OF VARIANCE BY RANKS WAY
y It is used when K samples are matched or dependent and are having ordinal data. y It is two - way analysis for differences. y It is a test that is very useful in determining whether K related samples are from the same population and hence, have no differences among themselves. y In it, the design has rows - representing set of matched subjects or respondents and column - representing various samples obtained under various conditions. y After presenting the data in a tabular form, each row scores are to be ranked.
ORDINAL DATA
MORE THAN TWO SAMPLES (RELATED)
(continued)
WAY
y If null hypothesis is true then the distribution of K ranks in each sample would be a matter of chance and hence, the sum of ranks for each column should be similar. y To test for the differences in the column totals of k sum, it makes use of the following statistics : 12 2 = (R 3 1 j 2 ) n ( k +) nk ( k + j = 1) 1
whe re k =num e of colum ; br ns n =num e of rows or num e of m br br atche sub cts; a d je nd Rj =sum ation of ranks in thejth sa ple m m .
Assume that a professor of management read somewhere that the time of the day can affect the students learning in the classroom. For that, he undertook an action research. He had selected 4 topics along with 4 quizzes related to each of them to be administered at the end of the lecture. The topics are selected randomly to be delivered at different times of the day followed by the related quiz. In a particular week, on Monday he had a lecture and quiz at 8:30 am; on Tuesday at 11: am; on Wednesday at 12:30 pm; and on Thursday at 2:30 pm. There were 19 students in the class; the grade points for each quiz was out of 5 points and the you wish toof What is alongthingthe time of the with grade points the students in each quiz administration were noted. PUT ON TEST?
Rns ak M n ak e Rn a 17 . 4 29 . 2 34 . 7 20 . 5
1 9 31 25 . 3 3 .0 0 0
K-S TEST
SIGN TEST
RELATED SAMPLES
UNRELATED SAMPLES
SIGN TEST
MEDIAN TEST
K-S TEST
MORE THAN 2 SAMPLES ORDINAL DATA RELATED SAMPLES THE FRIEDMAN TEST TWO WAY ANALYSIS OF VARIANCE BY RANKS UNRELATED SAMPLES
MEDIAN TEST
2 2 .2
2 2 .2
2 2 .2
S . D v= . td e Ma = en N = 22 . 22
2 2 . 2 2 . 2 2 . 2. 2 2 2. 2 2 2. 2 2 2. 2 2 2. 2 2 2. 2 2 2. 2 2 2. 2 2 2. 2 2 2. 2 2 2. 2 2
22 2
. 22 2
2 2 .2
T EP R E TB D F T H E CN OY A
df 35
1. What is Null and Alternative Hypothesis? 1. Should we accept Ho? 1. What would be the p-value if our test is one tail test?
2 2 2 2
Frequency
2 2 2 2 YS E N O
HS V I O T E I E L CD N R E ? A A I T R H S PA E A ODR S T
.2222222 .2222222
Understanding Output
CHI-SQUARE TEST
SAMPLE SIZE SAMPLE VARIANCE TEST VARIANCE TEST STATISTICS (CHI SQUARE) p-VALUE 1926 0.000281 0.000225 2402.202374 0.000000
What would be the research design? & What should be the appropriate test?
and variance = ( ( Di -
160.0000
10
66.0824
20.8971
Paire d Sample s Te st Paired Differences 95% Confidence Interval of the Difference Lower Upper
Mean Pair 1 SALES BEFORE THE INCENTIVE SCHEME (IN LAKHS) - SALES AFTER THE INCENTIVE SCHEME (IN LAKHS)
Std. Deviation
df
Sig. (2-tailed)
-5.0000
7.5277
2.3805
-10.3850
.3850
-2.100
.065
What would be the research design? & What should be the appropriate test?
S ig . .2 3 3
t .7 5 9 .7 5 9
df
What would be the research design? & What should be the appropriate test?
THE EQUALITY OF VARIANCE TEST FTEST H0 : 21 = 22 and H1 : 21 22 ; that is, it tests whether the samples are from two normal populations with equal variances. The test statistics used for it is - F = S21 /
S2
ANALYSIS OF
RELATED SAMPLES
UNRELATED SAMPLES
VARIATION TEST
PAIRED t-TEST
F-TEST