Hi Anna,
1. How does one interpret a strong correlation between a behaviour score and an LV brain score that does not have a low probability value under a 1000 permutations?
The permutation test is attempting to generate the expected distribution of parameters given a random sample and assuming the null hypothesis. With the number of voxels we have going into the analysis, it will often be the case that high correlations will occur by chance, which is what the permutation test trying to illustrate
2. What does the bootstrap scale represent on the LV brain maps? and what would be the lowest threshold you would tolerate for making pretty pictures for publications.
I can "roughly" be treated as a z-score - or better as an approximate confidence interval. I would use somewhere around 3 for most data sets to feel comfortable, which (very roughly) approximates a 99% confidence interval. Remember (as stated in the paper), that you are not making inferential statements with bootstrapping, that is what the permutation test is for.
Randy
Baycrest is an academic health sciences centre fully affiliated with the University of Toronto
Privacy Statement - Disclaimer - © 1989-2024 BAYCREST HEALTH SCIENCE. ALL RIGHTS RESERVED