I'm trying to run GPPPI on a dataset I previously analysed using SPM12.
I don't know if this is relevant but my first level SPM.mat files have more conditions than the four I specified in GPPPI and I also have a parametric modulator included in my first levels which I have not specified in the GPPPI.
I'm not getting an error message so I can't figure out where I could be going wrong.

Any help appreciated!

here's the P structure I set up:


%% specify ROI mask
P.VOI = 'K:\AAC\AAC_latest\analysis\second_level\ROIs_corrected\con_minus_noncon_mean_act_rACC.nii';

%% define region name
RegionName = 'con_minus_noncon_mean_act_rACC';

%% Creating a group folder
P.Region= RegionName;
P.GroupDir = ['K:\AAC\AAC_latest\analysis\PPI_group_level\' RegionName];

Group_Dir = P.GroupDir;
if ~exist(Group_Dir, 'dir')
mkdir(Group_Dir);
end

%% These are all PPI related variables that are identical for all subjects
P.Estimate=1;
P.contrast=0;
P.extract='eig';
P.Weights=[];
P.method='cond';
P.analysis='psy';
P.CompContrasts=1;
P.Weighted=0;
P.equalroi=0;
P.Tasks= {'0' 'ap_ap_decision' 'av_av_decision' 'con_ap_decision' 'con_av_decision'};

for i = 1:4
P.Contrasts(i).left=P.Tasks(i+1);
P.Contrasts(i).right={'None'};
P.Contrasts(i).STAT='T';
P.Contrasts(i).MinEvents=2;
P.Contrasts(i).name=P.Tasks(i+1);
end

P.Contrasts(5).left=P.Tasks(2);
P.Contrasts(5).right=P.Tasks(3);
P.Contrasts(5).STAT='T';
P.Contrasts(5).MinEvents=2;
P.Contrasts(5).name='ap_ap_minus_av_av';
P.Contrasts(6).left=P.Tasks(4);
P.Contrasts(6).right=P.Tasks(5);
P.Contrasts(6).STAT='T';
P.Contrasts(6).MinEvents=2;
P.Contrasts(6).name='con_ap_minus_con_av';

Hi Maria,

What exactly failed?

Your gPPI Model should have 8 condition columns (4 tasks, 4 pmod), 8 PPI columns 1 for each of the 8 conditions, seed region, and any covariates.

Best,
Donald

Hi Donald,
Thanks for your response.
This is the error I am getting.  I guess it means that the PPPI inputs are fine and it is some other error.
Any idea how I would troubleshoot?

Thanks,
Maria

ERROR 1: PPPI inputs were specified correctly, but failed error checking for dependencies. This is likely a bug in the program.
One or more inputs are not correct.

It looks like your missing P.directory in the specification of the P parameter structure. This should be the location of the first-level task SPM.mat file.

Can you set this and try again. If you still get an error, I'll have to take a closer look to identify which line might be causing an issue.

-Donald

Hi Donald,

I wanted to update that I resolved this issue.
The silly problem was that we had moved the preprocessed EPI files to another server (I had the P.directory specified but had missed it in my message).
There wasn't a specific error about not being able to find the file, which might be useful for a future update.

I have been writing some new contrasts to my PPI model (just using spm_FcUtil) and I wanted to ask about something.
When I ran the initial gPPPI the resultant con images are .img rather than .nii and when I try to reevaluate them using spm_contrasts I only get the con files for the ones which I have added (and not the ones which I also can see in the model, but were written by GPPPI).  SPM doesn't like these .img files for further analysis. Obviously I can just change the extension to .nii but I wanted to check that there isn't some reason to have .img files instead? Sorry if this is a silly question.

It all seems to work fine when I just rewrite the contrasts (which were initially written by GPPI) to the model and then run spm_contrasts.
Does this compromise the model at all? I was just unsure if there was something about the way that GPPPI sets up these contrasts. I had a look in both models and they seem identical (apart from SPM.xCon(1).Vspm.fname which is now a .nii file rather than an .img file). It seems to me all the gPPPI information is in the first level conditions and therefore I can treat these beta weights going forward as if they are a standard task based fMRI analysis.  Can you confirm?

Best,
Maria

I attached the original SPM design file to the last message and I attach the one with my rewritten contrasts to this one.

Thanks for the help!

The new SPM.mat file was not attached.

Ah. Yes. Moving the input files will cause an issue. I'll try and figure out how to make this error message clearer if this is the case in the next release.

Contrast Comments:
(1) spm_contrasts will default to only estimating contrasts that don't have a Vcon field. This is why it skips over existing contrasts.

(2) spm_contrasts for generating contrasts is the same as using spm_contrasts_ppi. There is no difference and it will not impact the model in anyway. The betas can be treated identically to how they are treated in a task-based fMRI analysis.

(3) nii versus img file naming is an issue with the script. I've attached a newer version called spm_contrasts_gPPI. If you rename this new file with spm_contrasts_ppi, and replace the old file, then the files should correctly have .nii extension again.

Let me know if you have any other issues. I hope to have a new python/SPM based package completed by this summer.

Thanks so much for the prompt response Donald! I really appreciate the help.

Seems like I don't need to send the SPM.mat file now as all is well.

Thanks again!