Covariate Selection for Group Comparison
Alan
Dear HCP Users,
I’m working with the 7T resting-state fMRI data (the file rfMRI_REST1_7T_PA_Atlas_MSMAll_hp2000_clean.dtseries.nii) from the Human Connectome Project, to perform a statistical comparison between two gender groups. For each subject, I compute a single value, and I then compare these values between males and females to assess statistical significance.
I would like advice on which covariates I should consider including in my statistical model to control for potential confounds. Age is can be a choice, but are there other variables I should include (e.g., motion parameters, cognitive scores, scanner-related variables, etc.) that are particularly relevant given the data I'm using?
Additionally, I appreciate guidance on where to find these required variables in the HCP dataset? I’ve looked into the behavioral CSV files, but I’d appreciate clarification on which specific files and variables would be most appropriate for this kind of analysis.
Thank you for your help.
Best regards,
Glasser, Matthew
I’m not sure there is a general answer to this question. We will shortly have some improved data available that will control for certain things (like global respiratory artifact) that may differ between males and females.
Matt.
--
You received this message because you are subscribed to the Google Groups "HCP-Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to
hcp-users+...@humanconnectome.org.
To view this discussion visit
https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/9280f49f-1dbb-4362-9527-88a0a587826fn%40humanconnectome.org.
The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.
Andraž Matkovič
subsection Methods / CCA modeling of many SMs and functional connectomes.
They used the following confounds:
1 Acquisition reconstruction software version (as an improved MRI reconstruction method was implemented in the third quarter of acquisition year 1).
2 A summary statistic quantifying average subject head motion during the resting-state fMRI acquisitions (this is the average, across all time points, of the time point-to-time point head motion, that measure being the linear distance moved, averaged across the head).
3 Weight.
4 Height.
5 Blood pressure – systolic.
6 Blood pressure – diastolic.
7 Hemoglobin A1C measured in blood.
8 The cube-root of total brain volume (including ventricles), as estimated by FreeSurfer.
9 The cube-root of total intracranial volume, as estimated by FreeSurfer.
In addition to identifying these nine confound SMs, we also demeaned and squared measures 2–9 (the first is a binary indicator), to create additional confound measures, to help account for potentially nonlinear effects of these confounds.
They do not explain why these confounding regressors were used. As Matt suggested, note that these confounds do not account for other artifacts that may noot be adequately cleaned without tICA.