To test the significance of the skewness difference between two distributions with $N_1$ and $N_2$ samples, would the following test work:

  1. Create a single array of all the samples from both distributions.

  2. Randomly shuffle the array .

  3. Split the array at index $N_1$.

  4. Calculate the $abs()$ of difference of the skewness/kurtosis statistics for the two partitions.

  5. Repeat until desired precision/out of unique permutations.

Then check if the actual $abs()$ difference of skewness/kurtosis statistics lie beyond the $p = 0.05$ interval.

  • 2
    $\begingroup$ I think this would work to test whether one distribution is significantly more skew than another. $\endgroup$ – Peter Flom - Reinstate Monica Oct 10 '12 at 23:07

Yes, this method is correct, assuming you do not care about the direction of the difference.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.