Mixed Block/Event First-Level Design in Nilearn

Hi there,

We’re attempting to model a mixed block design using Nilearn’s FirstLevelModel class (i.e., participants are presented with the same set of trial types - e.g., memory, perspective, valence, etc. - between different blocks); an example of our design matrix is below:

We’re running into issues in explicitly defining contrast weights after we run the first level GLM; Nilearn doesn’t seem to allow us to apply weight to the block-trial pair in concert. One option we’re considering is modeling the block-trial pair explicitly (e.g., high_trust-memory, high_trust-perspective, high_trust-valence, etc.).

Our question: Is this a sound approach to model this type of design? This would allow us to appropriately define our contrasts of interest BUT disregards the nested structure of the design.

Any suggestions would be appreciated!


I’m not sure what you mean with “apply weight to the block-trial pair in concert”. You means that you would like to compute the interaction between the two regressors ? If it is the case, I would model explicitly the interactions you’re interested in in the design matrix so that you can create contrasts focussing on these .
Does that address your concern ? Best,

That does address our concerns, thanks a lot for your response!