Nugent, Allison C. (NIH/NIMH) [E] wrote: > Hello, > SO, the problem wasn't ResMS. Examining the code for > spm_graph.m, around one 130, you can see that the Bcov reported to the > workspace is ResMS*SPM.xX.Bcov *only* when xSPM.STAT is a 'P' Other way round actually: line 142: if xSPM.STAT ~= 'P' translates as "if xSPM.STAT does not equal 'P'" So line 149: Bcov = ResMS*SPM.xX.Bcov; does get run for both T and F statistic images. > covariance calculated by SPM.xX.Bcov*ResMS is an order of magnitude > greater than the workspace Bcov following the plot function. Are you sure you're using the same ResMS as spm_graph, i.e. line 148: ResMS = spm_get_data(SPM.VResMS,XYZ); and at the same voxel? (maybe put a debugging break-point after lines 95-97 to check this) If so, then I can't see what could possibly be different, Bcov is not modified in spm_graph after line 149, so I expect spm_graph might be looking at a different voxel to you, maybe? Hope this helps, Ged.