A consulting firm wishes to evaluate the performance of four software packages (A, B, C, and D) as measured by the computational time. The experimenter wishes to control for two variables: the problem type and the operating system configuration used. Four classes of each of these two variables are identified. Each software package is used only once on each problem type and only once on each operating system configuration. Given table shows the computational times (ms) for the four software packages under these controlled conditions.
(a) Test at the 5% level of significance for the equality of the mean computational times of the four software packages.
(b) Find a 95% confidence interval for the mean computational time using software package C.
(c) Find a 95% confidence interval for the difference in the mean computational times of software packages A and C. Is there a difference between the two?
(d) Which software package would you choose and why?
(e) Was blocking of the two variables, problem type and operating system configuration, effective?
(f) If you had to rerun the experiment, what type of design would you use?
|
Problem Type
|
Operating System Configuration
|
|
I
|
II
|
III
|
IV
|
|
1
|
B 14.2
|
A 37.3
|
D 24.3
|
C 22.6
|
|
2
|
D 18.5
|
C 18.1
|
B 27.4
|
A 39.6
|
|
3
|
A 36.1
|
B 16.8
|
C 13.2
|
D 20.3
|
|
4
|
C 17.5
|
D 24.6
|
A 36.7
|
B 19.5
|