Hi, 

So, I'm trying to run gPPI and I'm stuck with that old classic of a "Missing Conditions" error when trying to estimate contrasts, suggesting I've got the naming wrong or something. Error pasted below.


VOI has 56 voxels in 2x2x2 space

VOI has 17 voxels in 3x3x3 space. This is in the the same space as the input data and functional mask.
SPM12: spm_fMRI_design (v5183) 12:55:26 - 05/07/2017
========================================================================
Saving fMRI design : ...SPM.mat saved
SPM12: spm_fmri_spm_ui (v6088) 12:55:27 - 05/07/2017
========================================================================
Mapping files : ...done
Calculating globals : ...done
Saving SPM configuration : ...SPM.mat saved
estimate_PPI.m
SPM12: spm_spm (v6842) 12:55:51 - 05/07/2017
========================================================================
SPM12: spm_est_non_sphericity (v6827) 12:55:55 - 05/07/2017
========================================================================
Estimation Failed
Warning: Missing conditions!!! Invalid Contrast
Invalid Contrast
PPI Contrasts were not estimated for some reason.


Thing is, I've been through everything with a comb, and there's no mismatch between the task names and contrast names or, well anything else. I'm pulling the names directly from the first level SPM design and the contrast names match.  I've even tried specifying a contrail (even though I'm not using derivatives), but to no avail. The error persists even when I specify a contrast like: 

P.Contrasts(i).left={'Neg_T'};
P.Contrasts(i).right={'none'};
P.Contrasts(i).STAT='T';
P.Contrasts(i).Weighted=0;
P.Contrasts(i).MinEvents=0;
P.Contrasts(i).name='Neg_T_BL';


This is really weird as I've basically just swapped the path names on a script that worked perfectly on another data set. 

Any pointers on what might be up are very much appreciated.

Best, 
Haakon

Hi Haakon,

The issue is that the model estimation step never completed. Does the test dataset provided with gPPI work properly with your version of SPM and Matlab?

It is possible that an update to SPM broke the gPPI code. A fix might be needed, but I would need to know what conditions cause it to fail.

Best,
Donald

Hi Don,

I figured it out. Well, at least I figured out how to fix it: In my previous scripts I'd run the Omnibus F-test. Setting P.contrasts to 0 fixed the problem.

Best,
Haakon

That is very odd. That setting should not have impacted the model estimation.

Can you send me your two scripts and I can try to recreate the error.