Can temporal correlation noise model be turned off in first level model GLM?

I’m analyzing a dataset where participants play a video game in the scanner, and I’m concerned that modeling the temporal correlation will remove some of my signal of interest. In nilearn’s first level GLM there’s the noise_model parameter that defaults to ar1. If I wanted to disable this feature can I do something akin to noise_model: None or do I need to do something else?

Hi @dlevitas, thanks for your question! At the moment you can only choose as input between an autoregressive model (lag-1 or higher order → 'ar<N>') or ordinary least squares ('ols') for time-independent noise modeling. See this example and explanation: Nilearn: Statistical Analysis for NeuroImaging in Python — Machine learning for NeuroImaging. I am not an expert on fMRI GLM analysis so I cannot expand further about this implementation. Perhaps someone else will chime in. If you think it is something that would be useful in general, you can also open an issue on the nilearn github and make a proposal for improving the class.

Got it, thanks @ymzayek