estimate_r1_glms: Estimate betas using Rank-1 GLM with separate designs...

In bbuchsbaum/fmrireg: R package for the anlysis of fmri data

Estimate betas using Rank-1 GLM with separate designs (Mumford-style approach)

Description

This function estimates single-trial (or single-condition) betas in a rank-1 framework by splitting the design matrix xdat$X into a list of separate sub-designs (X_list), one for each condition (or event). It then calls r1_glms_betas() to jointly estimate the HRF shape and the per-condition amplitudes.

Usage

estimate_r1_glms(dset, xdat, hrf_basis, hrf_ref, maxit = 100)

Arguments

dset	An fmri_dataset object containing voxel data and mask
xdat	A list typically returned by get_X(), containing: X: The stacked design matrix of dimension n \times (k*m) Base: The baseline design matrix, dimension n \times p
hrf_basis	A T \times m matrix of HRF basis functions
hrf_ref	A length-T vector representing a reference HRF shape
maxit	Maximum number of L-BFGS-B iterations (default 100)

Details

This is the “Rank-1 GLM with Separate Designs” approach (sometimes called the Mumford method). We partition xdat$X into k sub-designs, each n \times m. For each voxel, we remove the baseline using lsfit and call r1_glms_betas to solve for the event-wise betas under a rank-1 HRF constraint.

Value

A list containing:

beta_matrix	A k \times nvox matrix of estimated condition amplitudes
estimated_hrf	A T \times nvox matrix of the per-voxel estimated HRF shape

bbuchsbaum/fmrireg documentation built on March 1, 2025, 11:20 a.m.