Date: Feb 15, 2013 12:44 PM
Author: Luis A. Afonso
Subject: Re: Intra-permutations to test two different mean values
Follow-up

It was seen that the Hsu solution gave [1.720, 2.700] as the (central) 95% Confidence Interval (5% significance level ). It is interesting to check how no-parametric techniques behave before this spider?s problem. In this intention we performed:

__1) Bootstrap samples falling inside Hsu solution (2 experiments, 1 million each one): 0.97143, 0.97155.

__2) Intra-permutation: 0.97483, 0.97484.

Conclusion

In comparison with the Hsu test the no-parametric leads to lower the probability to reach significant values at this 5% level. To be close to this level Intra-permutations provides a slightly worse result than Bootstrap.

Luis A. Afonso