lyman.glm.build_design_matrix

lyman.glm.build_design_matrix(conditions=None, hrf_model=None, regressors=None, artifacts=None, n_tp=None, tr=1, res=60, shift=0.5, hpf_matrix=None, demean=True)

Use design information to build a matrix for a BOLD time series GLM.

Parameters

conditionsdataframe

Must have an onset (in seconds) columns. Can also have duration (in seconds, defaulting to 0), and value (in arbitrary units, defaulting to 1), and condition (strings, defaulting to “event”) columns; rows should correspond to event occurrences.

hrf_modelHRFModel object

Object that implements .transform() to return a basis set for the predicted response. Defaults to GammaHRF with default parameters.

regressorsdataframe

Additional columns to include in the design matrix without any transformation (aside from optional de-meaning). It must have an index with valid time points.

artifactsboolean series

A Series indicating which row should have indicator regressors included in the design matrix to account for signal artifacts.

n_tpint

The number of timepoints in the

trfloat

Time resolution of the output regressors, in seconds.

resfloat

Sampling resolution at which to construct the condition regressors and perform convolution with the HRF model.

shiftfloat

Proportion of the TR to shift the predicted response when downsampling to the output resolution.

hpf_matrixn_tp x n_tp array

Matrix for high-pass filtering the condition regressors.

demeanbool

If True, each column in the output matrix will be mean-centered.

Returns

Xdataframe

Design matrix with timepoints in rows and regressors in columns.