I performed a simulation study on R for bootstrap, permutation and classic version of hotelling T^2 test, Goodall F , and James T test
I aimed to compare 2 groups to see the performances of these tests for n= 20, 50, 100 and 500. For this reason I investigated type 1 error rates. For example I found that for n=20 type I error rate that belongs bootstrap version of hotelling tests is 0. How shall I interpret this result, I mean is it possible to say that the bootstrap version of hotelling test is unavailable for n=20...
As you can see the results below for Hotelling bootstrap version type 1 error rate is increasing together with the sample size... how can it be interpreted
thanx for your time
Uploaded with ImageShack.us
Mod Note: Would it be possible for you to crop your image so there isn't so much white space?
Last edited by Dason; 01-11-2011 at 09:24 AM.
Advertise on Talk Stats