tedana.metrics.dependence.calculate_semipartial_r2

calculate_semipartial_r2(*, data_optcom: ndarray, mixing: ndarray) → ndarray[source]

Calculate mean voxelwise semi-partial R-squared for each regressor.

Semi-partial R^2 is the incremental variance explained by adding a regressor to a model that already contains all other regressors.

We simplify the math by (1) orthogonalizing each component w.r.t. the other components then (2) calculating the R-squared for each component against the data.

Parameters:

  • data_optcom ((S x T) array_like) – Optimally combined data.

  • mixing ((T x C) array_like) – Mixing matrix.

Returns:

semi_partial_r2 ((C) array_like) – Average (across voxels) semi-partial R-squared for each regressor, on a scale from 0 to 100.