I have read a number of posts describing the issue around the large SnPM_ST file when wanting to do cluster inference testing.

I have two questions:

1. What are the pitfalls of changing the defaults threshold? Are you simply including fewer voxels?

This means that less information is saved in the SnPM_ST file. This results in a narrower range of cluster defining threshold you can choose in the Results step.

 

2. Is there a solution to this problem yet?

 

Specifying the cluster-defining threshold during the SetUp step can tremendously reduce the size of SnPM_ST. There are some ideas for eliminating the SnPM_ST file, but unfortunately we haven’t had a chance to implement it in SnPM yet. Using RANDOMISE in the FSL package is another possible solution.

 

-Satoru