I have run a searchlight analysis on 10 subjects and want to run a random effects analysis with mri_glmfit across subject. I have the results of each subject projected on the fsaverage surface. But I have no idea how to make it to 

1) compare the values in each voxel with 0.5 instead of 0

2) how to make it ignore the fact that I dont have variance for each subject and only have a single performance value (support vector machine performance) for each voxel (it looks like the mri_glmfit needs a ces.var.nii)

Any idea how I can do this? Sorry if this is a rookie question. I am totally new to searchlight and random effects analysis in fsfast. 

Thanks,

Maryam