Dear Donald, dear gPPI users,

I have successfully performed a gPPI analysis on my data a few months ago. I copied and pasted the results of the whole brain analysis on an excel sheet, and performed several beta extractions using Marsbar. I then re-opened the same second-level SPM.mat two weeks ago and found different results (i.e., the activated regions were relatively the same, but cluster size and significance were modified). Even worse, I re-launched the exact same gPPI and second level scripts to try and reproduce either the first or second set of data, and found... a third set of data.

Do you have any idea where these discrepancies could come from? For example, I was wondering if the gPPI could have modified the first level SPM.mat of the participants?

I would be very thankful for any input or help.

Kind regards,

Fabien Carruzzo

Hi,

I am completely unaware of how the second-level SPM.mat file could change as it is independent of the gPPI scripts. If you have the date of the original results, you could check if the SPM.mat file has been changed since then.

The gPPI script creates a separate SPM.mat in a new directory for the gPPI analysis, so the task-activity SPM.mat should not change. I'd have to check if the script overwrites existing SPM.mat files for the gPPI analysis.

I don't think any of the following will provide a solution, but they are worth checking:
(1) In the results that you are comparing, are the T values the same? This would indicate that the thresholds are different and not the underlying data.
(2) Check if the second-level SPM.mat files are different -->MATLAB has a function to compare mat-files. Once you identify the differences there, you can look at the source of those differences.
(3) If you have old and new 1st-level gPPI SPM.mat files, you could compare the SPM.mat files to see how they are different. The first place I would look is the design matrix.

Hope this helps,
Donald

Hi,

Thank you very much for your answer.

I performed the steps you indicated:
(1) The T values from the SPM.mat file that I copied and pasted and the ones that show up now are different, albeit pretty close. They are also different from the T values of the second gPPI analysis I made with the same scripts and data.
(2) And so, it's no surprise that the SPM.mat files from the two gPPI analyses are different. However, I do not have a version of the first SPM.mat I got to see what changed without doing anything with it.
(3) I took a look at the old first-level SPM.mat files, and they are identical to the ones I performed the gPPI analysis on. So, I shouldn't have got another set of values for the second gPPI. I will check again that my gPPI wrapper script is doing what it is supposed to do.

Kind regards,

Fabien