On Software Defect Prediction Using Machine Learning

Algorithm 3

Tukey’s multiple comparison: equal sample size.

Input: 𝛼, 𝑝, 𝑣, 𝑠, MSE, 𝑛_𝑡 and samples. The meaning of these parameters is: 𝛼 is an error rate.

𝑝 is the number of means, 𝑠=√MSE, 𝑣 is the degree of freedom related with MSE. 𝑛_𝑡 is the

number of observations of each sample.

Output: The minimum significant difference 𝜔 and a deduction;

Method:

(1) Choose a proper error rate 𝛼;

(2) Calculate the statistic 𝜔 according to

               𝜔=𝑞_𝛼𝑠(𝑝,𝑣)√𝑛_𝑡,

  where 𝑞_𝛼(𝑝,𝑣) is the critical value of Studentized range statistic, which can be found from any

  statistics textbooks.

(3) Compute and rank all the 𝑝 means

(4) Draw a deduction based on the ranks in the confidence level (1−𝛼).

End Algorithm Tukey Multiple Comparison.