Hi Anderson,

Thanks again! Very helpful! One last clarification:

>> Yes. I don't think it's necessary to run the full analysis, just the pre-stats, but doing all shouldn't hurt.

Ah OK, I assumed that I would need to run 1st Level analysis to account for Motion in the ROI time series. Is this not needed? I can just run Prestats to generate the time series, then run 1st Level with motion as a confound? (Looking back at the previous threads I cited, I believe this may align with Mark Jenkinson's recommendation. If so, makes things much easier for me.)

Warmly,

Erin

On Monday, July 18, 2016, Anderson M. Winkler <[log in to unmask]> wrote:

Hi Erin,

On 17 July 2016 at 23:24, Erin W. <[log in to unmask]> wrote:

Hi Anderson,

Thank you so much for the reply! A follow-up question: Can you elaborate on your response to this statement?

>> 2. What would you use as the INPUT for the second 1st Level analysis (which now includes PSYC, PHYS regressors)?

> The same as the earlier one, except that this time it would be for the full analysis, including smoothing, etc.

To ensure we are on the same page, I was assuming I would run the full 1st Level analysis two times (including prestats) to 1) Include Motion confounds (Standard Parameters & Outlier Matrix), and 2) To generate "filtered_func" to apply the ROI in subject space and obtain ROI time series (which accounts for motion). Let me know if this is incorrect. At least for the initial 1st Level analysis, this would be performed on "raw" (not pre-processed) data. Should I not be smoothing, etc, for the initial 1st Level analysis?

Yes. I don't think it's necessary to run the full analysis, just the pre-stats, but doing all shouldn't hurt.

 

Otherwise, I would assume the full 1st Level analysis is run twice; both times using "raw" data as the INPUT.

Yes. The 2nd analysis will be the true one, and will run in full. The 1st is only to compute the timeseries needed for the model used in the 2nd.

 

Both models would be identical, except the second time it is run, it would include PSYC and PHYS regressors?

Yes, so the models aren't identical, and in fact, no need for a model in the first run.

All the best,

Anderson

 

Thanks again for your input, much appreciated!

Warmly,
Erin

--
Sent from my iPhone