htfa.stats.gaussian_fit_statistics

htfa.stats.gaussian_fit_statistics(observed, predicted, *, n_parameters)

Score equally shaped (voxels, times) arrays or lists of subject arrays.

Supply the total nominal parameter count, INCLUDING one pooled residual variance. Likelihood uses sigma² = RSS/n and independent Gaussian residuals. Compare only the identical observations, masks, scaling and preprocessing. No intercept is fitted here; the temporal mean is only the R² baseline. Complex inputs and unrepresentable squared-error scales raise ValueError rather than silently discarding imaginary parts or reporting a perfect fit.