Hi,

I have noticed that when I set P.ConcatR=1 to concatenate across all runs when I do a gPPI analysis, that for the regressors that I don't
include in my contrasts, the corresponding PPI regressor has 0 variance/no "events"/"not unique" when reviewing the design matrix. I don't have this problem when P.ConcatR=0. I've attached the SPM.mat for a random subject for which I specified a minimum of 9 events for the contrast. Below is how I specified the contrasts.

Perhaps I missed something in the gPPI config? Thanks in advance!
Gina


conditions = {'conCinsTcnfrem0' 'conCinsTcnfrem1'...
'conCinsDcnfrem0' 'conCinsDcnfrem1' 'conIinsTcnfrem0' 'conIinsTcnfrem1'...
'conIinsDcnfrem0' 'conIinsDcnfrem1'};

P.ConcatR = 1; % use concatenated approach to deal with low # trials (<10 events)/event type
P.wb = 0;         % 0=slice-by-slice computations rather than whole brain -- this is already set by default

P.CompContrasts = 1; % estimate contrasts

for c=1:length(conditions)
     conname = conditions{c};
     P.Contrasts(c).name = conname; % gppi auto prepends con_PPI to name
     P.Contrasts(c).left = {conname};
     P.Contrasts(c).right = {''};
     P.Contrasts(c).STAT = 'T';
     P.Contrasts(c).MinEvents = 9;
end

You only selected a subset of tasks for the initial PPI model. The concatR=1 option uses all the tasks in the SPM model for the concatenated PPI model. However, the concatenation feature pulls from the initial PPI model.

The way the code is written is that it builds a PPI column for each task based on the task list from the task analysis SPM model, then it goes to find the PPI term for that task in the initial PPI model. As it can't find that PPI term as its not in the initial model, the column is set to all 0s.

In some future version of the code, there will be a change that will stop this from happening. However, its not high on the priority list as it has minimal effects.

Two solutions:
(1) ignore this issue as it won't impact the results using standard analytical methods; or
(2) specify all conditions in the P.Tasks parameter setting.

My personal opinion is to specify all conditions in the P.Tasks parameter setting.