You are on page 1of 8

CASE STUDYNO.

(TWO-SAMPLE TESTS: POOLED-VARIANCE T TEST FOR THE

DIFFERENCE BETWEEN TWO MEANS)

The highest paid instructor of CTU-DC has been teaching Engineering Economics for

20 years. He observed that experimental method is effective than conventional method

in teaching Engineering Economics. To prove this observation/claimed, I conduct a

study of a new method of teaching Engineering Economics, students were divided at

random into two sections A and B, of sizes 40 and 35 respectively. Those in section A

were taught using the experimental method, while those in section B were taught by

conventional methods. After completion of the course all the students were given the

same test paper, and the section scores were tabulated on the next page.

Is there any evidence that experimental method is effective?


SCORE TABULATION

Note: Data Were Collected from Engineering Students at The Cebu Technological University-Danao Campus.
SOLUTION:

Step 1: State the null and alternative hypothesis.

The null hypothesis will be that the two methods are equally effective, we have

𝐻𝑜 : 𝜇𝐴 = 𝜇𝐵 . Since B is an experimental method a natural alternative to consider is

whether it improves on the conventional method, A, so we choose the one-sided

alternative hypothesis 𝐻1 : 𝜇𝐴 < 𝜇𝐵 .

Step 2: Choose the level of significance and the sample size.

𝛼 = 0.01 𝑎𝑛𝑑 𝑛 = 75.

Step 3: Select the appropriate test statistic.

Since the population variances 𝜎𝐴 2 and 𝜎𝐵 2 are unknown, a natural test to consider

is a two-sample t–test, which depends for its validity on equality of these two variances.

In this case 𝑆𝐴 2 and 𝑆𝐵 2 are very close, so such an assumption is very reasonable.

The appropriate test statistic for the hypothesis 𝐻𝑜 is then,

̅̅̅1 − 𝑋
(𝑋 ̅̅̅2 ) − (𝜇2 − 𝜇1 )
𝑡𝑆𝑇𝐴𝑇 =
1 1
𝑆𝑝 √𝑛 + 𝑛
1 2

Where

(𝑛1 − 1)𝑆12 + (𝑛2 − 1)𝑆22


𝑆𝑝 = √
𝑛1 + 𝑛2 − 2

𝑆𝑝 =Pooled Variance

̅̅̅1=Mean of the sample taken from population 1


𝑋
𝑆12 =Variance of the sample taken from population 1

𝑛1 =Size of the sample taken from population 1

̅̅̅2=Mean of the sample taken from population 2


𝑋

𝑆22 =Variance of the sample taken from population 2

𝑛2 =Size of the sample taken from population 2

𝑡𝑆𝑇𝐴𝑇 =Test statistics follows a t distribution at 𝑛1 + 𝑛2 − 2 degrees of freedom

Step 4: Determine the rejection region.

We use a natural notation, with a subscript A or B indicating the group concerned. For

a given sample size, n, the test statistic 𝑡𝑆𝑇𝐴𝑇 follows at distribution with 𝑛𝐴 + 𝑛𝐵 −

2 degrees of freedom. The critical values of the t distribution with 35 + 40 –2 = 73

degrees of freedom are solve using t table and MS Excel. The alternative hypothesis,

𝐻1 : 𝜇𝐴 = 𝜇𝐵 , has one tail. If the alternative hypothesis includes the less than sign, the

critical value of t is negative. As shown in Table E.3 and Figure 9.11, because the

entire rejection region is in the lower tail of the t distribution and contains an area of

0.01, due to the symmetry of the t distribution, the critical value of the t test statistic

with 35 + 40 –2 = 73 degrees of freedom is 2.3785.

The decision rule is:

Reject 𝐻𝑜 if 𝑡𝑆𝑇𝐴𝑇 < −2.3785, otherwise do not Reject 𝐻𝑜 .

Solving the critical values using t table:

From the t table as shown below, the critical value is 2.3785.


Source: Extracted from Table E.3 of Basic Business Statistics: Concepts and Applications, 12th

Ed. by Berenson, Levine, and Krehbiel.

Alternate solution in solving the critical value using MS Excel:

Degrees of Freedom =𝑛𝐴 + 𝑛𝐵 − 2 =35 + 40 − 2 = 73

Lower Critical Value: =-TINV (2*0.01,73)

Result: ≈ -2.3785
Step 5: Collect the data and compute the value of the test statistic.

From MS Excel:

Group A
No. Name of Student Scores Group B
1 ABANES, JUNERICH R. 70 No. Name of Student Scores
2 ABOGADIE, JUN MAR M. 55 1 ABANILLA, RAYMOND G. 71
3 ACIDELLA, CHRISTIAN JOHN A. 40 2 ARCUINO, JAN ARVIN T. 54
4 ALVARADO, DEE R. 40 3 ARREZA, JOHN ARVI O. 59
5 ARAR, JAVIER D. 40 4 BARING, RYAN Q. 59
6 ARCILLA, REMZON C. 72 5 BAROSA, JASON M. 64
7 ATIZON, EUTIQUIO R., JR. 51 6 BATING, JONJIE J. 69
8 BERNAL, HARVEY F. 61 7 BINUEZA, SEAN ROVI DOMINIC T. 80
9 CACANOG, JOHN ALBERT A. 95 8 CAÑETE, JULITO I. 81
10 DEJAÑO, SHELA MARIE P. 91 9 CASUNO, ARTEMIO A. 56
11 DIMPAL, HAROLD T. 71 10 CENIZA, DANILO STEPHEN G. 47
12 ELI, ARSENIO P., JR. 47 11 CLAROS, JONEL B. 83
13 ESPINOSA, JUNFLOYD G. 74 12 CONIZA, REYMART S. 72
14 FIEL, ENMAR JOHN B. 49 13 CORROS, KALVIN JOSEPH T. 61
15 GABUNILAS, LEMUEL R. 52 14 DEJITO, ROGELIO Y. 85
16 GARCES, MARK RANDALE C. 54 15 DELA CRUZ, FRANCIS BONG B. 83
17 GONZALES, JENNY ANN G. 62 16 DENOY, JEREMIAH L. 75
18 GULFAN, SHALOM A. 88 17 DUAY, JON ERNEST M. 92
19 HIYAS, LEYSANDER RAMSE M. 63 18 ESTUDILLO, ELINO AUGUSTUS L. 72
20 LAPINID, REGINE O. 75 19 FABROS, VANESSA Y. 92
21 LIPARANON, ROY ADRIAN V. 60 20 GALABIT, NICOLAS N. 63
22 MANZANES, JEFFREY G. 92 21 GALANO, ALFIO S. 73
23 MENDOZA, JUMIL A. 41 22 GALANO, RHEY JOHN B. 77
24 NOYA, JOHNES C. 93 23 GARNICA, JAYSON C. 75
25 PANILAG, LOLIBERTH L. 70 24 GARRIDO, MAUREEN JOY C. 91
26 PIAMONTE, JACOB HOPE E. 59 25 GONTIÑAS, FELIZARDO A., JR. 60
27 SINANGOTE, JAPHET LUKE C. 89 26 GONZALES, JAYCE N. 49
28 TANEO, JONATHAN D. 60 27 LINGANAY, ELMER E. 77
29 TORREGOSA, JORAM CESLY 80 28 MAHAYAG, ARCHIE A. 95
30 TORSINO, MARK JAYSON C. 82 29 MERIN, CLARK DHARRYL L. 90
31 TUDIO, JERICK JOVIN R. 69 30 OBQUIA, VINCENT ANGELO B. 71
32 VELASQUEZ, JESON B. 90 31 OJEDA, MARVEY E. 70
33 MONTON, WINSTON MARL R. 40 32 ROCHE, EVAMIE P. 92
34 PABILAN, NEIL JOHN C. 42 33 RONQUIZ, HAZEL B. 82
35 PADILLA, RENE BOY B. 50 34 ROSAL, GLADYS ANN S. 90
36 PEREZ, CHRISTIAN M. 50 35 ROSIANA, GLENDO L. 70
37 GORILLO, ALWENNE L. 43
38 LAGUNA, JESSIE JAMES S. 36
39 MASONG, BON CARL DOMINIC T. 35
40 HEMOTA, JESSA DIANE B. 83
Pooled-Variance t Test for the Difference Between Two means
(assumes equal population variances)
Data
Hypothesized Difference 0
Level of Significance 0.01
Population 1 Sample
Sample Size 40 =IF(COUNTA(Table5[Name of Student])=0,"",COUNTA(Table5[Name of Student]))
Sample Mean 62.8500 =AVERAGE(Table5[Scores])
Sample Standard Deviation 18.5120 =STDEV(Table5[Scores])
Population 2 Sample
Sample Size 35 =IF(COUNTA(Table8[Name of Student])=0,"",COUNTA(Table8[Name of Student]))
Sample Mean 73.7143 =AVERAGE(Table8[Scores])
Sample Standard Deviation 13.1296 =STDEV(Table8[Scores])

Intermediate Calculations
Population 1 Sample Degrees of Freedom 39 =K7-1
Population 2 Sample Degrees of Freedom 34 =K11-1
Total Degrees of Freedom 73 =SUM(K16:K17)
Pooled Variance 263.3732 =((K16*K9^2)+(K17*K13^2))/K18
Standard Error 3.7562 =SQRT(K19*(1/K7+1/K11))
Difference in Sample Means -10.8643 =K8-K12
t Test Statistic -2.8923 =(K21-K4)/K20

One-tail test
Lower Critical Value -2.3785 =-TINV(2*K5,K18)
p -value 0.0025 =IF(K22<0,K30,K31)
Reject the null hypothesis =IF(K26<K5,"Reject the null hypothesis","Do not reject the null hypothesis")

Note:
Cell F31 0.002518038 =TDIST(ABS(K22),K18,1)
Cell F32 0.997481962 =1-K30

Alternate solution in solving 𝑡𝑆𝑇𝐴𝑇 :

(𝑛1 − 1)𝑆12 + (𝑛2 − 1)𝑆22


𝑆𝑝 = √
𝑛1 + 𝑛2 − 2

(𝑛𝐴 − 1)𝑆𝐴2 + (𝑛𝐵 − 1)𝑆𝐵2


𝑆𝑝 = √
𝑛𝐴 + 𝑛𝐵 − 2

(40 − 1)(18.51202 ) + (35 − 1)(18.51202 )


𝑆𝑝 = √
40 + 35 − 2

𝑆𝑝 =16.22878
̅̅̅1 − 𝑋
(𝑋 ̅̅̅2 ) − (𝜇1 − 𝜇2 )
𝑡𝑆𝑇𝐴𝑇 =
1 1
𝑆𝑝 √𝑛 + 𝑛
1 2

̅̅̅
(𝑋 ̅̅̅̅
𝐴 − 𝑋𝐵 ) − (𝜇𝐴 − 𝜇𝐵 )
𝑡𝑆𝑇𝐴𝑇 =
1 1
𝑆𝑝 √𝑛 + 𝑛
𝐴 𝐵

(62.8500 − 73.7143) − 0
𝑡𝑆𝑇𝐴𝑇 =
1 1
16.22878√40 +
35
𝑡𝑆𝑇𝐴𝑇 = −2.8923
Step 6: State the statistical decision and the managerial conclusion.
Because 𝑡𝑆𝑇𝐴𝑇 = −2.8923 < −2.3785,We therefore reject the null hypothesis at the

1% level of significance, and conclude that there is strong evidence that method B

improve on its conventional method.

(See also the solutions of Case Study No. 2 using MS Excel on the accompanied MS
Excel file of this document.)

You might also like