作者: Johann Bacher
关键词: Probabilistic clustering 、 Mathematics 、 Statistics 、 Model selection 、 Latent class model 、 Probabilistic latent semantic analysis 、 Probabilistic relevance model 、 Statistical hypothesis testing 、 Local independence 、 Latent variable model
摘要: This paper develops a probabilistic clustering model for mixeddata. The allows analysis of variables mixed type: thevariables may be nominal, ordinal and/or quantitative. modelcontains the well-known models latent class as submodels.As in analysis, local independence isassumed. parameters are estimated by EMalgorithm. Test statistics and goodness-of-fit measures proposedfor selection. Two artificial data sets show usefulness ofthese tests. An empirical example completes presentation.