Ben,
See my inline responses below.
On Thu, Oct 9, 2014 at 3:24 PM, Ben Gold <[log in to unmask]> wrote:
> Hi SPMers,
>
> I'm working with the gPPI toolbox to contrast the contribution of my seed
> region in two groups defined by their psychological conditions: in other
> words, I want to analyze the same regressors in each subject, but contrast
> them at the second level
>
gPPI is tailored specifically to doing what you want.
>
> My first question is why I'm getting an error message while doing this.
> The error occurs in the second spm_spm step, when saving planeEstimation
> fails (at plane 14/68, block 1/1). Then it says "PPI contrasts were not
> estimated for some reason." Has anyone else encountered this, or does
> anyone know what the problem might be?
>
It could be a memory issue. Do you have enough free space on the drive?
Does it always stop on the same plane?
>
> My second question, which might help answer the first one, is about my
> parameter structure. In order to identify the regressors I'm interested in,
> I'm using the 'cond' method and specifying the tasks with their names
> (which are 'fix,' 'punishment,' and 'reward'). Then to create t-contasts to
> simply select the regressors of interest (which I in turn plan to contrast
> at the second level), I have:
>
> P.Estimate=1;
>
> P.Tasks={'1' 'fix' 'punishment' 'reward'};
>
> P.CompContrasts=1;
>
> P.Contrasts(1).left={'fix' 'punishment' 'reward'};
>
> P.Contrasts(1). right={'none'};
>
> P.Contrasts(1).STAT='T';
>
> P.Contrasts(1).MinEvents=5;
>
> P.Contrasts(1).name='Music';
>
>
> To my understanding, this should do what I described above: select the
> regressors I'm interested in and create a contrast that includes all of
> them for me to use at the second level. Am I wrong, and/or is this causing
> the problem with estimating the PPI contrasts?
>
The estimation error above is the estimation of the model, not the
contrasts. I should change the error message to say "PPI contrasts were not
computed...." Without having a complete and estimated model, the contrast
(you only specified 1) can not be computed.
The way the contrasts are set up is based on the null hypothesis. In your
contrast, the null hypothesis was:
Ho: (fix+punishment+reward)/3=0
Then the left side of the equal sign goes into the left field, and the
right side goes into the right field.
I suspect that you want 3 contrasts, 1 for fix, 1 for punishment, and 1 for
reward. If this is the case, then you need 3 Contrasts.
P.Contrasts(1)
P.Contrasts(2)
P.Contrasts(3)
The name and left fields should change for each contrast to reflect the
condition being contrasts against baseline.
Hope this helps.
>
> Thank you for your help!
>
> -Ben
>
> --
> Ben Gold
> Doctoral Candidate
> Integrated Program in Neuroscience
> Montreal Neurological Institute
> McGill University, Canada
>
