jmcm News

CHANGES IN VERSION 0.2.5

USER-VISIBLE CHANGES

• improve readability of the code.

• add some useful comments.

• speedup MCD/ACD/HPC method by paralleling code of objective function / gradient computation using OpenMP.

CHANGES IN VERSION 0.2.4

USER-VISIBLE CHANGES

• improve readability of the code.

• add some useful comments.

• speedup MCD/ACD/HPC method (GH #2, @Neutron3529).

CHANGES IN VERSION 0.2.3

BUG FIXES

• fix the warning on order() suggested by Kurt Hornik.

• remove lattice from Imports.

• fix a bug on model order (GH #1).

USER-VISIBLE CHANGES

• speedup HPC method (GH #2, @Neutron3529).

CHANGES IN VERSION 0.2.2

BUG FIXES

• fix function bootcurve(): argument optim.method was missing to call function optimizeJmcm(), and now it is set to 'default'.

CHANGES IN VERSION 0.2.1

USER-VISIBLE CHANGES

• add the JSS paper as vignette in package jmcm.

CHANGES IN VERSION 0.2.0

USER-VISIBLE CHANGES

• add option optim.method in function jmcm(): by default, optim.method is 'default' so that the BFGS algorithm we implemented internally is used; alternatively, we can set optim.method as 'BFGS' so that vmmin() (the underlying C function used in stats::optim() for BFGS algorithm) will be called through R package roptim.

NEW FEATURES

• class JmcmBase is implemented as the base class for class MCD, ACD and HPC to reduce code duplication.

• class JmcmFit is used for fitting of the three Cholesky-based joint mean-covariance models (i.e., MCD, ACD and HPC).