statsmodels.robust.covariance.CovDetMCD

class statsmodels.robust.covariance.CovDetMCD(data)

Minimum covariance determinant estimator with deterministic starts

Preliminary version.

Parameters:

dataarray_like

Multivariate data set with observation in rows and variables in columns.

Methods

fit(h, *[, h_start, mean_func, scale_func, ...])

Compute minimum covariance determinant estimate of mean and covariance

Notes

Reproducibility: this uses deterministic starting sets and there is no randomness in the estimator. However, this will not be reproducible across statsmodels versions when the methods for starting sets or tuning parameters for the optimization change.

The correction to the scale to take account of trimming in the reweighting estimator is based on the chisquare tail probability. This differs from CovMcd in R which uses the observed fraction of observations above the metric trimming threshold.

References

..[1] Hubert, Mia, Peter Rousseeuw, Dina Vanpaemel, and Tim Verdonck. 2015.

“The DetS and DetMM Estimators for Multivariate Location and Scatter.” Computational Statistics & Data Analysis 81 (January): 64-75. https://doi.org/10.1016/j.csda.2014.07.013.

..[2] Hubert, Mia, Peter J. Rousseeuw, and Tim Verdonck. 2012. “A

Deterministic Algorithm for Robust Location and Scatter.” Journal of Computational and Graphical Statistics 21 (3): 618-37. https://doi.org/10.1080/10618600.2012.672100.

Methods

fit(h, *[, h_start, mean_func, scale_func, ...])

Compute minimum covariance determinant estimate of mean and covariance