Robust and Unbiased Variance of GLM Coefficients for Misspecified Autocorrelation and Hemodynamic Response Models in fMRI
Figure 8
Ratios of estimated and true contrast variance for the event-related and blocked design when both the correlation structure and HRF model are incorrect. Two cuts of both the sandwich (blue) and smoothing approach (red) variance estimates are shown, at and for event-related, and at and for blocked design.
