Browse Forums

help

47 Subscribers

help > gPPI -SPM estimation fails - please help!

Showing 1-1 of 1 posts

gPPI -SPM estimation fails - please help!

Hi Dr McLaren (and everyone),

Thank you for providing the gPPI toolbox.

I have an event-related task with 4 conditions, where each trial randomly presented and is separated with an intertrial interval (rest/fixation-cross) period. The 1^st level model space has conditions specified as the regressors of interest, and noise regressors of no interest.

I have successfully managed to run the example data upon downloading the toolbox and changing a couple of the scripts with the scripts I collected from the posts in this forum. However, when I run the toolbox on my data, I ran into errors. I managed to fix some of them but the issue still continues.

In running the data, I used the following P structure;

P.subject: '152'

P.directory: '/path/to/subject/1st_level'

P.VOI:'/path/to/subject /PPI_1st/masks/vmPFC_mask.nii'

P.Region: 'vmPFC'

P.Estimate: 1

P.extract: 'eig'

P.Tasks: {'0' 'CondA' 'CondB' 'CondC' 'CondD'}

P.Weights: []

P.maskdir: '/path/to/subject/1st_level'

P.equalroi: 0

P.FLmask: 0

P.analysis: 'psy'

P.method: 'cond'

P.CompContrasts: 0

P.Contrasts(1).name='PPI_CondB_CondA

P.Contrasts(1).left='CondB'

P.Contrasts(1).right='CondA'

P.Contrasts(1).MinEvents=20

P.Contrasts(1).STAT='T'

P.Contrasts(2).name='PPI_CondC_CondA'

P.Contrasts(2).left='CondC'

P.Contrasts(2).right='CondA'

P.Contrasts(2).MinEvents=20

P.Contrasts(2).STAT='T'

P.Contrasts(3).name='PPI_CondD_CondA'

P.Contrasts(3).left='CondD'

P.Contrasts(3).right='CondA'

P.Contrasts(3).MinEvents=20

1) When I run gPPI on my data, I got the output below, then estimation failed. It creates SPM.mat file with missing fields. Looking at the SPM fields created I realized the toolbox seemed not to extract the timeseries from the mask (Timeseries regressors for the mask is all NaN).

2) Adding P.preservevarcorr=1 to the ‘P’ parameter structure (preserving variance correction estimated from the 1^st level), lets the estimation go through a little bit more, but the estimation still gets stuck at the estimation of spatial non-sphericity.

I am not sure whether estimation fails because of a problem with the time-series extraction or the timeseries extraction fails because of the estimation problem.

Any advice, insight and help would be greatly appreciated!!

###################################

Valid Contrast