statsmodels.robust.covariance.CovM.fit#

CovM.fit(start_mean=None, start_shape=None, start_scale=None, maxiter=100, update_scale=True)[source]#

Estimate mean, shape and scale parameters with MM-estimator

Parameters:

start_meanNone or ndarray: Starting value for mean, center. If None, then median is used.
start_shapeNone or 2-dim ndarray: Starting value of shape matrix, i.e., scatter matrix normalized to det(scatter) = 1. If None, then scaled covariance matrix of data is used.
start_scaleNone or float: Starting value of scale.
maxiterint: Maximum number of iterations.
update_scalebool: If update_scale is False, then the scale is fixed at start_scale and only mean and shape are updated in each iteration.

Returns:

CovMResult: Named tuple with mean, shape, scale, cov, converged, n_iter, and mahalanobis. See CovMResult for details.

Notes

If start_scale is provided and update_scale is False, then this is an M-estimator with a predetermined scale as used in the second stage of an MM-estimator.