[MATH-1240] Kolmogorov-Smirnov 2-sample test returns wrong p-value if input data is identical

• Type: Sub-task
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 4.0, 3.6
• Assignee: Unassigned
• Reporter: Thomas Neidhart
• Created: 24/Jun/15 11:47
• Updated: 25/Jan/16 20:27
• Resolved: 28/Jun/15 10:00

Description

Depending on the method used, wrong results are returned:

• exact -> correct
• monteCarlo -> wrong (returns a very small p-value)
• approximate -> wrong (returns 0 instead of 1)

The problem for the approximate version is due to this code snippet in ksSum

        if (t == 0.0) {
            return 1.0;
        }

should rather return 0 in this case, but needs to be verified.

This problem was hidden (due to wrong statistic calculation) in previous versions of commons-math or resulted in a TooManyIterationsException (if statistic is 0 as correct in case of identical values).