for my dissertation I need to compare the Gini coefficients of 5 different markets with approx. the sample population size. I would like to do bootstrapping to compare the means (instead of permutation tests). What I've currently done is:

1. I bootstrapped all markets individually 1000 times and received 1000 Gini-coefficients each, so 5000 in total for 5 markets.

2. I used the 1000 Gini values in a two-group t-test to compare their mean with another 1000 Gini-values from a different market and thus received a p-value.

Observation: Even small differences tend to become significant.

Is this a potential way to do it? I've read other board entries on this, but I still don't quite get it what to do. Thanks for your help!


  • $\begingroup$ The problem of small differences being significant can be addessed by reporting effect size, or by reformulating your hypothesis to be more precise, rather than just testing for any effect. But there is also a problem of repeated testing: why do several (10?) t-tests instead of a multi-group testing? Also, is there a published justification for (1)? It may not be a defensible approach - you are attempting to generate independent variables by subsampling the same distribution. But I admit the context is not clear to me. And they likely converge much sooner than 1000 (?) $\endgroup$ – katya Dec 15 '14 at 18:09

Your Answer

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

Browse other questions tagged or ask your own question.