Package: pGME
Type: Package
Title: Estimates the parameters of normal mixtures by penalized
        likelihood methods (with exact Newton's method in multivariate
        case).
Version: 1.0
Date: 2012-08-10
Author: Grigory Alexandrovich, Florian Schwaiger
Maintainer: Grigory Alexandrovich
        <alexandrovich@mathematik.uni-marburg.de>
Description: Estimating the parameters of normal mixtures can lead to difficulties (especially for small sample sizes), if each component of the mixture has possibly a different mean and standard deviation resp. covariance matrix, since then the likelihood is unbounded for any standard deviation parameter going to zero. Further, when estimating a mixture with too many components the parameters are not identifiable anymore since weight parameters can converge towards zero. Both mentioned cases can be avoided by using penalty functions in the log-likelihood.
License: GPL-2
Depends: methods,mclust,mvtnorm,rgl,ellipse
Packaged: 2013-08-29 14:30:42 UTC; schwaige
