New methods OPP and POP for the estimation of nu: nu_OPP_estimator() and nu_POP_estimator().
Function fit_mvt() updated using POP as default method for nu.
Vignette updated.
Functions fit_Tyler() and fit_Cauchy() now recover the missing scaling factor with the improved OPP-harmonic method.
New contributors added for OPP and POP methods: Frederic Pascal and Esa Ollila.
New method for skewed t distributions: fit_mvst()
New contributor added for fit_mvst(): Xiwen Wang.
fit_mvt() and fit_mvst(): Now the bounds for nu estimation can be set as a global option, e.g.: options(nu_min = 4.2).
Fixed description regarding covariance matrix for Cauchy distribution.
fit_mvt(): It accepts weights as argument to weight differently the samples (as opposed to uniform weights).
fit_mvt(): Many more methods to estimate nu iteratively (via argument nu_iterative_method).
fit_mvt(): New argument scale_covmat to include a correction factor in the covariance matrix for minimum MSE (still in development).
Vignette revised: detailed descriptions of the algorithms included.
Comparison with additional existing benchmark sn::selm() included in the vignette.
Now the three fitting functions also return the number of iterations and elapsed cpu_time.
Significant revision of the fitting function fit_mvt(); in particular:
Function fit_mvt() now allows the choice (via the argument na_rm) to drop the observations with NAs
or impute them.
Initial release is on CRAN.
It includes three functions for heavy tails fitting: fit_mvt(), fit_Tyler(), and fit_Cauchy().
Vignette illustrates its use and comparison with existing packages.
Tests are included.
fit_mvt() can deal with NAs and a factor model structure on the covariance matrix.