topic Re: Non-parametric 2 sample t-test in Statistical Procedures

Non-parametric 2 sample t-test

Melk — Mon, 07 Aug 2017 06:25:57 GMT

I want to test the difference in "central tendency" - for lack of a better word- of 2 samples where 1 is realtively normal looking and the other is highly skewed to the right. Normally, I would just use a wilcoxon rank sum test, but I read that the 2 distributions must be similar in shape. Since mine are not, what test is appropriate?

Re: Non-parametric 2 sample t-test

Ksharp — Mon, 07 Aug 2017 12:25:44 GMT

Since wilcoxon rank sum test is a non-parameter test. It doesn't matter what distribution your data have.

wilcoxon rank sum test can handle any distribution. Or you would like compare the median of them ?

I remember @Rick_SAS wrote a bolg about it before .

Re: Non-parametric 2 sample t-test

Rick_SAS — Mon, 07 Aug 2017 12:58:17 GMT

I agree with KSharp: it sounds like you can use PROC NPAR1WAY for this analysis.

The blog post he refers to shows that you can use PROC QUANTREG for a two-sample test of the difference between arbitrary quantiles. It can also provide confidence intervals for the quantiles.