Significance Tests & Confidence Intervals for Non-normal data

183 Views Asked by At

I have 2 nonnormally distributed samples of different sizes (N1~=N2).

To evaluate whether there is a significant difference between these samples, I used the Mann_whitney U test (ranksum in MATLAB)

Now I want to evaluate by how much the populations differ. With normally distributed data, I would just use the difference of the means with a SEM confidence interval.

However, for nonparamteric data, my wikipedia research suggested using the Hodges–Lehmann estimator. My questions are:

1) is the Hodges–Lehmann estimator indeed correct

2) how do I calculate confidence intervals for the HL estimator (like the SEM)

3) How can I implement this in matlab?

Thanks