Does nilearn have the option to compute weighted matrices as CONN is able to do? The CONN documentation says that “In pure resting-state analyses, weights are defined to encompass entire runs”. Is it possible to do this with nilearn on rest data?

I don’t think it has an option to do that. however as conn seems to compute a simple linear regression and scikit-learn’s linear regression accepts sample weights, you could use a nilearn masker to extract time series, nilearn glm compute regressor to convolve the weights with the HRF, and scikit-learn’s linear regression to solve the least-squares problem