Munich Personal RePEc Archive

Assessing the number of components in a normal mixture: an alternative approach

Maciejowska, Katarzyna (2013): Assessing the number of components in a normal mixture: an alternative approach.

[img]
Preview
PDF
MPRA_paper_50303.pdf

Download (267kB) | Preview

Abstract

In this article, a new approach for model specification is proposed. The method allows to choose the correct order of a mixture model by testing, if a particular mixture component is significant. The hypotheses are set in a new way, in order to avoid identification problems, which are typical for mixture models. If some of the parameters are known, the distribution of the LR statistic is Chi2, with the degrees of freedom depending on the number of components and the number of parameters in each component. The advantage of the new approach is its simplicity and computational feasibility.

UB_LMU-Logo
MPRA is a RePEc service hosted by
the Munich University Library in Germany.