Hello Gabor,
Thank you for your upload - for your two-sample paired t-test since you have repeated measures you need to supply an exchangeability-blocks file to randomise ( with the -e option ). With this option I would expect the same number of permutations to be generated in both cases.
Many Regards
Matthew
>
>> Hi all,
>>
>> I have 7 subjects, each measured twice (two time points). I'd like to do a nonparametric two-sample paired t-test using randomise.
>>
>> It is really interesting if I subtract the two timepoints manually for each subject and do a one sample non-parametric t-test on the 7 difference images it leads different result than doing a two-sample paired t-test on the original images (14 images). The design for paired t-test was made using Glm, and the subjects for the two-sample paired t-test were given in the right order into the 4D file. It is also interesting for me, that if I do the paired t-test it is possible to run the analysis with n=500 permutations as well. I think that analyzing the difference images using one-sample t-test should be equivalent with paired t-test, therefore I don't understand why randomise is able to run 500 permutations and not just 2^7=128.
>>
>> Does anybody have any idea about this issue? Maybe I missed some basic thing?
>>
>> Thanks a lot,
>> Gábor
>>
>
|