3 ms·
Agree with you on all three, but specifically for 1., can you think of pathological pairs of distinct distribution that the test would often fail to reject?
by murbard2 11y ago
Agree with you on all three, but specifically for 1., can you think of pathological pairs of distinct distribution that the test would often fail to reject?
- tgb 11y agoThe article says it's poor at detecting differences in the tails and much better at differences in the medians. So that's where I'd start to find problems.
- murbard2 11y agoPlaying with the tails make all kind of mistakes possible, but that seems like a criticism that would apply to any attempt to identify a distribution based on sample.