I have two lists. Each of list can have integers 0, 1, 2, etc. I want to test if the distribution is significantly different from each other. Is it correct to use a Kolmogorov-Smirnow test or some other non-parameteric test?
$\begingroup$
$\endgroup$
2
-
1$\begingroup$ Can the integers go to infinity, or just to 2? Where do these integers come from? What do you mean if the distributions are "significantly different from each other"? Do you just want to know if the means differ, if the SDs do, if they differ in skewness, all 3? Etc. $\endgroup$– gung - Reinstate MonicaCommented Mar 26, 2016 at 2:05
-
$\begingroup$ Hi, The integers can go up to some finite number but usually somewhere around 4 to 10 at most. Another challenge is that more than 50% of the numbers are simply 0, so median gives me 0. I want to know if their means different (and I have been using Mann Whitney U test, but possibly other tests are also possible) $\endgroup$– Lee SandeCommented Mar 26, 2016 at 23:05
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
You say
The integers can go up to some finite number but usually somewhere around 4 to 10 at most.
so with a low upper limit you can assemble your data as a contingency table and do a chi-square test of homogeneity. That would be for an all-inclusive null hypothesis of equal distributions.
But you also say
I want to know if their means different
and one simple way of doing that is a permutation test, implemented via simulation.
answered Mar 8, 2021 at 7:39