Dear Dr. McLaren,

I has hoping I could seek your help on the following two questions that relate to gPPI:

1) When specifying 'tasks' in the parameter structure, do you only need to specify regressors which are relevant to the experiment? e.g. do you also need to account for time and dispersion derivatives? 

2) If one wishes to look for a difference in connectivity between conditions, then from my understanding you simply specify P.conrast.left (-) P.contrast_right. However what if you are specifically interested in the parametric modulators (PMs) that are assigned to those conditions? From what I read on another help topic, the gPPI method also accounts for the PM's assigned to tasks, but I'm wondering how to test for a difference in task-related connectivity between conditions, versus a difference in the way connectivity is modulated by the relevant PM between conditions.

Many thanks,

ME

Marcos,

1) When specifying 'tasks' in the parameter structure, do you only need to
specify regressors which are relevant to the experiment? e.g. do you also
need to account for time and dispersion derivatives?

>>>The time and dispersion derivatives are not computed for the interaction
terms; but they are included for the task effects. The reason for this is
that we are assuming a fixed HRF for all brain regions and thus there is no
delay or dispersion of the PPI effect. The main reason for including the
time and dispersion derivatives is because the event duration is unknown.
The PPI effect - by definition - is known because you specify when the the
neural events occur.


2) If one wishes to look for a difference in connectivity between
conditions, then from my understanding you simply specify P.conrast.left
(-) P.contrast_right. However what if you are specifically interested in
the parametric modulators (PMs) that are assigned to those conditions? From
what I read on another help topic, the gPPI method also accounts for the
PM's assigned to tasks, but I'm wondering how to test for a difference in
task-related connectivity between conditions, versus a difference in the
way connectivity is modulated by the relevant PM between conditions.

>>>PM are computed for the PPI terms if they were included for the task
regressors. To compute the difference in PMs for PPI, set the contrail
field to the PM tail that is in SPM.xX.name (e.g. xtime^1). The setting
would then be entered as: P.Contrasts(ii).Contrail={'xtime^1'};

>>>To test the effect of the PPI task regressors, without PM, just leave
the Contrail field blank.

>>>I would not test the PPI for the task versus the PPI for the PM as the
question does not make sense. I would test each separately using 2
contrasts.

Hope this helps.

Dear Dr. McLaren,

Many thanks for your reply, it was indeed very helpful.

I had one follow-up question.

Suppose you want to look for a difference in PMs for PPI between several conditions. For example, let's say one has a hypothesis that connectivity is modulated as a function of PM for condition A and B, but not C. Then I assume you would put A and B on the left hand side of the contrast equation and C on the right, and look for a positive PPI effect. However, I wonder how one might infer the directionality of any subsequent PPI effect, i.e. whether connectivity goes up or down as the PM goes up or down for A and B?

I hope this makes sense.

Very best wishes,

Marcos

Please see inline comments below.

---

Suppose you want to look for a difference in PMs for PPI between several
conditions. For example, let's say one has a hypothesis that connectivity
is modulated as a function of PM for condition A and B, but not C. Then I
assume you would put A and B on the left hand side of the contrast equation
and C on the right, and look for a positive PPI effect. However, I wonder
how one might infer the directionality of any subsequent PPI effect, i.e.
whether connectivity goes up or down as the PM goes up or down for A and B?

>>> Put condition A on the left hand side and 'None' on the right hand
side. This will compare condition A against 0 and will tell you if it goes
up or down. If the contrail is the PM, then the contrast will test if the
condition A PM is greater than or less than 0. Hope this helps. NOTE: SPM
only displays positive values, so if you want to review the negatives in
SPM you'd need to create the reverse contrast. If you are just using PMs as
inputs into a second-level analysis, you only need one direction as the map
will include both positive and negatives.

Hope this helps.

Put condition A on the left hand side and 'None' on the right hand
side. If the contrail is the PM, then the contrast will test if the
condition A PM is greater than or less than 0. Hope this helps.

>>> Just for absolutely clarification, a positive PPI resulting from condition A PM on the left and 0 on the right (with contrail set to PM) is interpreted as greater connectivity with increasing PM values? And a negative PPI would be the opposite? Can you specify multiple conditions on the left hand side and 0 on the right to look for an average difference from 0?  

NOTE: SPM only displays positive values, so if you want to review the negatives in
SPM you'd need to create the reverse contrast. 

>>> I'm presuming you are referring to first-level maps only here?

Your responses have been extremely helpful.

Marcos

Just for absolutely clarification, a positive PPI resulting from condition
A PM on the left and 0 on the right (with contrail set to PM) is
interpreted as greater connectivity with increasing PM values?

>>> Correct.

And a negative PPI would be the opposite?

>>> Correct.

Can you specify multiple conditions on the left hand side and 0 on the
right to look for an average difference from 0?

>>> Correct. I would be careful about doing this though. You need to make
sure that the PMs are scaled in a similar manner. For example, if one PM is
trial number and the other PM is post-scan rating, then I wouldn't average
those because they are scaled differently.


NOTE: SPM only displays positive values, so if you want to review the
negatives in
SPM you'd need to create the reverse contrast.

>>> This is for any contrast in SPM - not specific to PPI. If you want the
negative for the second level, you have to build a negative contrast.

Dear Dr. McLaren,

I apologize for posting yet another question, however I thought this issue was worth mentioning.

I am trying to conduct a simple PPI to look for a difference between two conditions. However, in my GLM I have included a particular regressor of onsets twice, each time with a different PM attached, i.e. Onsets A (PM 1) Onsets A (PM 2). This is because I did not want to orthogonalize the PMs but rather force them to compete for variance.

When I run PPPI it crashes with the following message:

Estimation Failed
Warning: Missing conditions!!! Invalid Contrast
Invalid Contrast
Warning: Missing conditions!!! Invalid Contrast
Invalid Contrast
PPI Contrasts were not estimated for some reason.

When I checked the SPM file, I realized a PPI regressor for Onsets A and PM 1 had been created twice, but no PPI regressor for PM 2. I'm guessing this is the problem.

I know that my GLM setup is unusual, but there is any way around this, other than removing the repeated onset regressor? 

Many thanks,
Marcos

The task names need to be unique. This means that having the same task name
repeated in the model will cause problems for gPPI.

The solution is to make the task names unique (e.g. OnsetsA1 and OnsetsA2).

Dear Mr. McLaren,
I am new to PPI Analyses and I am very sorry for bothering you with this somehow unspecific question.
I have an event-related design with 11 different conditions. There are approximately 8 events per condition.
I set up a wrapper script and the P structure for the PPPI function looks just fine, however I get an error such as

Invalid Contrast
Warning: Missing conditions!!! Invalid Contrast

This is puzzling, as the conditions can be seen at P.Tasks and they look just fine.
The error is created when the new regressors are created (in a script called createvec).
However I don't understand the underlying problem.
Could you or anybody else give me any suggestions on what to change?

Thank you very much and best from Berlin
Miriam