Dear Jeff and Hannes,

You are right that the original QA motion and global change second-level covariates computed automatically during preprocessing are computed by aggregating across all timepoints (across multiple sessions and multiple conditions). If you want to compute these separately for each condition you can simply use the 'covariate tools. compute summary measures' function to do this. The QA_timeseries first-level covariate contains the global change and framewise displacement timeseries so it is just a matter of computing the aggregated measure that you wish for each of these timeseries (e.g. average or maximum value across timepoints) and making sure to select the 'condition-specific measures' checkbox so that the measures are computed separately for each condition (e.g. go to 'covariates.first-level' tab, select the 'QA_timeseries' covariate, click on 'covariate tools.compute aggregate measures', then select 'raw values', 'average', and 'do not aggregate', check the 'condition-specific measures' and click 'Ok').

If you want to compute the measures separately for each session simply make sure that you define a new set of conditions first where each condition looks at the data of a single-session (e.g. see the 'pre-post' example in the manual for how to define conditions associated with individual sessions).  

Perhaps it is a good idea to try to change the default behavior and have CONN create condition-specific QA measures also during preprocessing? (let me know your thoughts; by default the QA measures created during denoising are computed separately for each condition, the reason that the ones generated during processing are not is that we cannot safely assume that conditions have been yet defined at the time when people run preprocessing)

Hope this helps
Alfonso
 
Originally posted by Jeff Browndyke: