You are on page 1of 1

1 Technically, one should run a two-sample hypothesis test for differences in proportions.

However, Excel does not provide this tool,


and consequently it is not discussed in the book. So what should a student do? Here is a chance to build upon previous knowledge to
apply confidence intervals. Using the concepts from the previous chapter (see page 192 on CIs and decision making), we could construct
confidence intervals for the data and compare them. For example, for Dealer Satisfaction, all confidence intervals overlap,
For example, for Dealer Satisfaction, all confidence intervals overlap, although the overlap between NA and SA, and SA and China is slight,
suggesting that there might be a significant difference between them.
Similarly, it appears that there is a difference between SA and China in End User Satisfaction also.
Dealer Satisfaction
Survey Scale:
North America
2012
South America
2012
Europe
2012
Pacific Rim
2012
China
2012
End-User Satisfaction
North America
2012
South America
2012
Europe
2012
Pacific Rim
2012
China
2012

15

44

56

125

22

60

5 Sample
Size

90

0.91

0.0300

0.85

0.97

17

30

0.83

0.0680

0.7

0.96

12

0.75

0.1250

0.51

0.99

16

0.63

0.1210

0.39

0.87

0
0

1
2

2
3

3
15

4
31

5 Size
49

100

0.80

0.0400

0.72

0.87

19

37

37

100

0.74

0.0439

0.65

0.83

19

45

33

100

0.78

0.0414

0.7

0.86

19

42

35

100

0.77

0.0421

0.69

0.85

31

14

50

0.90

0.0424

0.82

0.98

Top Box ProportionsStandard error Lower 95% CI Upper 95% CI


0.80
0.0358
0.73
0.87

2
Using ANOVA, we reject the null hypothesis that all ratings are the same; so at least one differs from the rest.

Anova: Single Factor


SUMMARY
Groups

Count

Quality
Ease of Use
Price
Service

Sum

200
200
200
200

Average

879
833
734
828

4.395
4.165
3.67
4.14

55.505
621.65

3
796

18.50167
0.780967

677.155

799

Variance

0.581884
0.610829
1.136784
0.794372

ANOVA
Source of Variation

SS

Between Groups
Within Groups
Total

df

MS

23.6907

P-value

F crit

1.1E-14 2.616089

3
The proportion of on time deliveries in 2008 was 0.9850.
We may test the null hypothesis that the proportion of on time deliveries in 2012 is < 0.985 to determine if it has improved (the alternate hypothesis is p > 0.985)
The sample proportion for 2012 is 0.9907
z = (0.9907 - 0.985)/SQRT(.985*(1-0.985)) =
0.046893
Critical value = 1.645
p-value =
0.48129912
Therefore, we cannot conclude a significant improvement
4 A chart of the average number of defects by year shows a declining trend.
1000
800
600
400

200
0
2008

2009

2010

2011

2012

We may test for differences between 2008 and 2012 (assuming the samples are the monthly data since we don't know the actual number of shipments)

t-Test: Two-Sample Assuming Unequal Variances


2008

Mean
826.3333333
Variance
135.3333333
Observations
12
Hypothesized Mean Difference
0
df
12
t Stat
20.61894864
P(T<=t) one-tail
4.88441E-11
t Critical one-tail
1.782287556
P(T<=t) two-tail
9.76883E-11
t Critical two-tail
2.17881283

2012

496.25
2940.022727
12

The test clearly shows a significant difference in the mean defect rates.
5 Testing hypotheses that the mean cost has improved for one of the new processes, we cannot conclude a signficant improvement.

t-Test: Two-Sample Assuming Unequal Variances


Current

Mean
289.6
Variance
2061.144828
Observations
30
Hypothesized Mean Difference
0
df
52
t Stat
0.283404509
P(T<=t) one-tail
0.388996025
t Critical one-tail
1.674689154
P(T<=t) two-tail
0.77799205
t Critical two-tail
2.006646805

Process A

285.5
4217.637931
30

t-Test: Two-Sample Assuming Unequal Variances


Current

Mean
289.6
Variance
2061.144828
Observations
30
Hypothesized Mean Difference
0
df
41
t Stat
-0.968320801
P(T<=t) one-tail
0.169280943
t Critical one-tail
1.682878002
P(T<=t) two-tail
0.338561887
t Critical two-tail
2.01954097

Process B

298.4333333
435.3574713
30

6 Conduct two sample tests on mean years at PLE for each factor.

t-Test: Two-Sample Assuming Unequal Variances


Female

Mean
5.530769231
Variance
12.25064103
Observations
13
Hypothesized Mean Difference
0
df
18
t Stat
-0.009174759
P(T<=t) one-tail
0.496390313
t Critical one-tail
1.734063607
P(T<=t) two-tail
0.992780627
t Critical two-tail
2.10092204

Male

5.540740741
6.449430199
27

CONCLUSION: NO SIGNIFICANT DIFFERENCE BY GENDER

t-Test: Two-Sample Assuming Unequal Variances


College Grad N

Mean
4.892307692
Variance
5.819102564
Observations
13
Hypothesized Mean Difference
0
df
29
t Stat
-1.078815194
P(T<=t) one-tail
0.144780886
t Critical one-tail
1.699127027
P(T<=t) two-tail
0.289561772
t Critical two-tail
2.045229642

College Grad Y

5.848148148
9.10951567
27

CONCLUSION: NO SIGNIFICANT DIFFERENCE BY COLLEGE GRAD STATUS

t-Test: Two-Sample Assuming Unequal Variances


Local N

Mean
3.629411765
Variance
5.690955882
Observations
17
Hypothesized Mean Difference
0
df
34
t Stat
-4.418641425
P(T<=t) one-tail
4.80512E-05
t Critical one-tail
1.690924255
P(T<=t) two-tail
9.61023E-05
t Critical two-tail
2.032244509

Local Y

6.947826087
5.272608696
23

CONCLUSION: SIGNIFICANT DIFFERENCE BY LOCAL AREA


EMPLOYEES FROM THE LOCAL AREA HAVE GREATER RETENTION

You might also like