Hi Anderson, 

Apologies - I would like to investigate whether the amount of modulation in significant clusters corresponds to clinical and cognitive variables in my patient sample. 

I'm not exactly sure what values to extract from significant clusters. Would it be parameter estimates from the modulated design? Or maybe percent signal change values from the non-modulated model (that models each load with a separate regressor)?

Thank you in advance for the advice!
Lara

On Sat, Jan 20, 2018 at 6:05 AM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Lara,

Sorry, I don't quite understand what you'd like to do. So you have a significant group difference for the parametric regressor, and you now have the %BOLD change for each subject in that significant region. What exactly would you like to investigate?

All the best,

Anderson


On 19 January 2018 at 17:41, Lara Foland-Ross <[log in to unmask]> wrote:
PS - I should mention that I extracted % signal change values from separate 1st level feat analyses that modeled each working memory load using separate regressors and no parametric modulation. 

Thanks in advance for the help!

On Fri, Jan 19, 2018 at 2:28 PM, Lara Foland-Ross <[log in to unmask]> wrote:
Hi Anderson, 

Thanks for your help. Using this approach, I have obtained some interesting differences at the group level, between patients and controls. I'd now appreciate some guidance regarding how best to extract signal change from significant clusters.

What I've done thus far is use significant clusters as ROIs to extract percent signal change from each of the 3 working memory load conditions (separately) for every participant, for every ROI. So mean activation for 0-back, mean activation for 1-back, and mean activation for 2-back, for each roi and for every subject.

To land on an estimate of how much a subject modulates their activation in a given ROI, would I then calculate the slope?

Thanks again,
Lara


On Mon, Dec 25, 2017 at 8:34 PM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Lara,

Yes, that looks right.

All the best,

Anderson


On 22 December 2017 at 16:28, Lara Foland-Ross <[log in to unmask]> wrote:
Hello FSLers,

I'm setting up a GLM that uses parametric modulation for the first time and would appreciate confirmation that I am on the right track. My task involves 3 working memory load conditions (blocked), as well as instruction blocks and a resting (blank screen) baseline.

It's my understanding that I should model three regressors:

EV1: instruction cue condition (no parametric modulation; 3rd column of input text file = 1)
EV2: working memory condition w/no parametric modulation (3rd column of input text file = 1; no discrimination between 3 WM loads and all WM events modeled together)
EV3: working memory condition w/parametric modulation; (3rd column of input text file = mean-centered WM load)

I then create the following contrast to estimate the modulation of WM on activation:
0 0 1

Is this setup is appropriate?

Many thanks in advance,
Lara