2ndlevelmodel covariates / nuisance regressors contrast help

Hi All,

I’m having a bit of confusion while working with nistats.second_level_model.SecondLevelModel

Originally I was just using an intercept column as my design matrix, but our lab is now trying to include age, sex, and mean framewise displacement (meanFD) of the scan as covariates (somewhat mimicking the OASIS example).

Here is an example of what the Covariates dataframe looks like.

For each task, only the FD for that task is kept, subjects with NaNs for the task are dropped (e.g. 8 out of 100), an intercept column is added, and the contrast is [0, 0, 0, 1] (weighting the intercept and using Age, Sex, and task_meanFD as covariates).

Originally, our 2ndlevel maps had many large clusters with high magnitudes that survived thresholding, but running 2ndlevels with this more detailed design matrix reduced our thresholded maps to essentially no activation.

I’m not sure if the issue is with the scale of the meanFD columns or the contrast, but something seems incorrect. Any guidance would be greatly appreciated!