3dttest covariates

AFNI Message Board
1

Hi,

I have the following question.

I have data from CONN (denoised whole brain time series) and would like to account for age and gender, and I used 3dttest, but it only takes 3d images.

I figured you can use -brickwise and -covariates at the same time, but -resid cannot be used with brickwise. Is there another option to do this: account for age and gender on a time series.

Best,
Omar

2

Omar,

I’m confused with your description. Are you talking about voxel-wise time series for each subject? Is this a subject- or population-level analysis? With the former, how do gender and age come into play? With the latter, you would have to elaborate it more. What is exactly your model?

3

Gang Wrote:

Omar,

I’m confused with your description. Are you
talking about voxel-wise time series for each
subject? Is this a subject- or population-level
analysis? With the former, how do gender and age
come into play? With the latter, you would have to
elaborate it more. What is exactly your model?

I am also a little confused too since my boss asked me to do this.

Yes, it is a voxel-wise time series (no stats calculated yet) for each subject.

This would be a subject-level analysis. What they basically want is to account for age and gender on each voxel-wise map, if that makes sense in order to account for those effects in the time series for each subject.

We have 2 groups: healthy control and psychosis.

We want to use that data for subsequent analysis like machine learning and not have to account for age and gender in those models.

Hope this makes sense.

4

If this is just a subject-level analysis, it does not make sense to consider gender and age in the time series regression model because both are FIXED for each each subject. Only at the population level would it become reasonable to incorporate the two variables. And 3dDeconvolve/3dREMLfit, not 3dttest++, would be the appropriate program in your case.