Minimum covariance determinant estimator#
- MCD(data, h, seed=None, mfull=10, nstep=7, hiRegimeCompleteLastComp=True)[source]#
- Description
Calculates the Minimum Covariance Determinant covariance matrix
- Arguments
- data
Matrix of data where each row contains a d-variate point.
- h
Size of the data subset to use during estimation.
- mfull
In the high regime n>600, number of best results we keep before computing on the full dataset (cf paper by Rousseuw and van Driessen).
- hiRegimeCompleteLastComp
“True” if in the high n regime case in the last computation we carry computation until convergence of the solutions, false if we use a fix amount of nstep number of steps.
- nstep
In high n regime, finite number of steps to carry last computations for final solutions if we do not want to compute until convergence (hiRegimeCompleteLastComp is set to false).
- References
Peter J. Rousseeuw & Katrien Van Driessen (1999) A Fast Algorithm for the Minimum Covariance Determinant Estimator, Technometrics, 41:3, 212-223