Building GLM with missing conditions in some subjects

I am trying to build first level GLMs in SPM for a memory task which has 3 stimuli types (X, Y, and Z) and two possible response types for each condition (“A” and “B”). I want to have regressors for each response type and stimuli (so six total conditions to model), however there are some conditions that never occur in some subjects (e.g. some subjects never response “A” to Z, others never respond “B” to X).

Is it okay if some subjects’ first level models just don’t include regressors for conditions that never happened for them? If not, how could I include that regressor if SPM does not allow empty values for onset and duration of conditions?

Any insights would be appreciated! Thank you!

1 Like

I have once had the same problem and I was generating sparse random stimulus onsets for such conditions. Unfortunately, this lowers your sensitivity a bit.


Wiktor Olszowy