Hi all,
I’m working with an fMRI dataset where each session is intended to have two runs, but some sessions are missing one run.
When combining runs at the session level using FSL’s fixed-effects model, I’ve been assigning weights as follows:
- If both runs are present: each run gets 0.5
- If only one run is present: that run gets 1.0
This way, each session contributes equally regardless of missing data.
In one example (see screenshot below), the second session includes only a single run, so I assigned it a weight of 1.0. Sessions with two runs each had weights of 0.5 per run.
I’ve applied a similar strategy when combining sessions into subjects, to ensure that participants with fewer sessions don’t contribute disproportionately to the group-level analysis.
I’d really appreciate any clarification on whether this kind of manual weighting is necessary:
- Does FSL internally account for the number of runs/sessions per subject in fixed-effects or group-level models (e.g., FLAME)?
- Or would omitting this adjustment risk introducing bias into the estimates?
Many thanks in advance for your help and insights!