mne.stats.linear_regression: can it be used to predict an ERP response

💬 Support & Discussions
, , ,
1
  • MNE version: 1.7.0
  • operating system: macOS-12.6

Hi all,

I am using mne.stats.linear_regression to calculate regressed ERPs against one regressor (“stimulus time”, the time that a stimulus appears in each trial). Now I want to get a predicted ERP for a stimulus time value that hasn’t appeared in the design matrix “stimulus time” regressor. Will it work if I multiply the beta associated with the “stimulus time” regressor resulting from mne.stats.linear_regression and multiply it with a new design matrix that contain the new “stimulus time” regressor values. Namely,
predicted_erp = np.tensordot(new_design_matrix.values, np.stack([beta_regressor, beta_intercept], axis=0), axes=([1], [0]))

Appreciate any advice and help!

Thanks,
Lizzie