Negative Correlation in Whole Brain

[z641509906]

I am new in fmri research. I have encountered same situation(Same as issue 300). Does anyone know what caused it?
I am trying to use my data in Searchlight for RSA demo. The data used are the beta values from GLM which have 4 run and 3 conditions. The model used is a 3 * 3 (1-Identity Matrix) repeated 16 times(4 run * 4 run) which means that the beta of same condition in different runs are similar.
Problem: When I use Identity Matrix as RDM instead of (1-Identity Matrix), there are brain regions showing positive correlation. However, when I used (1-Identity Matrix) as RDM, the whole brain shows negative correlation. The results of demo are successfully repeated.
Trying:

1. I have tried different methods to calculate correlation and scores, which are not helpful.
   Possible Reasons:
2. Preprocessing? I preprocess the bold data using SPM following Andy's Brain Book(https://andysbrainbook.readthedocs.io/en/latest/index.html). Maybe the procedure will introduce correlation between voxels?
3. Group Analysis? I tried do group analysis among 8 subjects. I have done ttest for each seachlights and plotted the tmap. Maybe ttest is not reasonable?
   Thanks a lot for any suggestion.

[JasperVanDenBosch]

hi @z641509906 , I'm moving this to the discussion forum since this is more of a methods question than a technical issue with the toolbox.