I have performed the KS test comparing two samples. My sample size is very large so the p value is very low even though the effect looks weak. I found a an explanation for the calculation of the effect size here (How do I calculate the effect size for the Kolmogorov-Smirnov Z statistic?) which suggests dividing the KS statistic by the $\sqrt\frac{n1*n2}{n1+n2}$. Are there any other methods for calculating effect size? I saw here (Robust effect sizes for 2 independent groups) that the KS statistics itself can be used directly as the effect size.

Secondly, is there a rule of thumb as to what is considered a big or small effect size in this case?

  • 2
    $\begingroup$ How you assess the effect depends on why you are performing this test. In the majority of applications described on this site, the test is irrelevant or superfluous, so this is an important consideration. What is the reason you are applying it? $\endgroup$
    – whuber
    Commented Aug 22, 2018 at 13:31


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.