statsmodels.regression.mixed_linear_model.MixedLM.get_scale#
- MixedLM.get_scale(fe_params, cov_re, vcomp)[source]#
Return the estimated error variance based on given estimates of the slopes and random effects covariance matrix
- Parameters:
- fe_paramsarray_like
The regression slope estimates
- cov_re2d array_like
Estimate of the random effects covariance matrix
- vcomparray_like
Estimate of the variance components
- Returns:
- scale
float The estimated error variance.
- scale
Notes
If the random effects covariance matrix is singular or a variance component is zero, then directions with a variance below 1e-10 are dropped from the covariance of the observations, which is the limit as these variances go to zero.