Negative Data

[smeisler]

Hello,

I would like to use XCP_D (https://github.com/PennLINC/xcp_d) to post-process my CIFTI data from fMRIPrep (as opposed to doing confound regression here). During confound correction, I use global signal regression, alongside other regressors, which effectively centers my data around 0. Therefore, a lot of vertices have low or negative means. I was wondering if the meanTol uses the absolute value of vertices or also considers the sign. I am guessing that it considers the sign since when I use my XCP_D outputs, over half of the vertices are excluded due to the low means check. If this is the case, would it be possible to include negative values with magnitude above the mean tolerance, or would that cause issues with modeling? I hope my question made sense, and happy to provide more info as needed.

Thanks, Steven

[mandymejia]

Hi Steven,

Thanks for bringing this up, and pardon the slow reply. Damon is working on adding an alternative method of scaling the data, and building in an automatic check for near-zero means to select the best scaling method. Instead of dividing by the mean, we would divide by the standard deviation, like a z-score. The interpretation of the task activation coefficients would be the change in BOLD signal in units of standard deviations that happens due to the task. Currently the interpretation is the percent BOLD signal change that happens due to the task. There is also the alternative of not scaling at all, but then the coefficients would be more difficult to interpret.

Damon, can you weigh in on the timeline for the change?

Best, Mandy

On Fri, Feb 11, 2022 at 5:50 PM Steven Meisler @.***> wrote:

Hello,

I use XCP_D (https://github.com/PennLINC/xcp_d) to post-process my CIFTI data from fMRIPrep. During confound correction, I use global signal regression, alongside other regressors, which effectively centers my data around 0. Therefore, a lot of vertices have low or negative means. I was wondering if the meanTol uses the absolute value of vertices or also considers the sign. I am guessing that it considers the sign since when I use my XCP_D outputs, over half of the vertices are excluded due to low means. If this is the case, would it be possible to include negative values with magnitude above the mean tolerance, or what that cause issues with modeling? I hope my question made sense, and happy to provide more info as needed.

Thanks, Steven

— Reply to this email directly, view it on GitHub https://github.com/mandymejia/BayesfMRI/issues/22, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABHVKURKQG2LBOXE2UPZ33TU2WHCLANCNFSM5OFRZV2A . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

You are receiving this because you are subscribed to this thread.Message ID: @.***>

-- AMANDA F MEJIA, PhD Assistant Professor Department of Statistics Indiana University https://mandymejia.com/

[damondpham]

Hello Steven,

I'll aim to address this and your other Issue by Thursday of next week!

[damondpham]

Hi Steven,

I addressed this (and your other Issue) just now. Try the newest version of 1.9 and let me know how it goes!