I don't understand the main objective to use these statistical techniques and how with they could compares an algorithm with other.
You might want to investigate whether Rank selection was faster at solving a problem compared to Tournament selection. Or you might want to investigate whether PMX Crossover was more efficient at solving a problem compared to Ordered Type 1 Crossover.
You could do some experiments by finding out how many generations or iterations are needed to solve the problem and you might get data like so.
Run generations needed to solve generations needed to solve using Tournament Selection using Rank selection
1 1432 564
2 587 444
3 998 668
. . .
25 1265 842
In order to find out if there is a difference in the number of iterations needed to solve the problem you might do a t test on the iterations you obtained for multiple runs. You would do an F test to see if the variances are equal or not. You will choose an appropriate t test depending on whether the variances are equal or not.
see
http://www.statsdirect.com/help/default.htm#parametric_methods/unpaired_t.htm
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques