# model.selection.R: Selection of both the number K of clusters and the subset S... In ClustMMDD: Variable Selection in Clustering by Mixture Models for Discrete Data

## Description

The inference on both the number K of clusters and the subset S of clustering variables is seen as a model selection problem. Each competing model is characterized by one value of ≤ft(K,S\right). The competing models are compared using penalized criteria AIC, BIC, ICL and a more general penalized criterion with a penalty function on the form

pen≤ft(K,S\right)=α*λ*dim≤ft(K,S\right),

where

• λ is a parameter that can be calibrated using "slope-heuristics" (see backward.explorer, dimJump.R),

• and α is a coefficient in [1.5, 2] to be given by the user.

## Usage

 1 2 model.selection.R(fileOrData, cte = as.double(1), alpha = as.double(2.0), header = TRUE, lines = integer()) 

## Arguments

 fileOrData A character string or a data frame (see backward.explorer). If fileOrData is a data frame, it must contains a column named logLik and another named dim (see details). cte A penalty function parameter. The associated criterion is -log(likelihood)+cte*dim. alpha A coefficient in [1.5,2]. The default value is 2. header Indication of the presence of header in the file. lines A vector of integer. If not empty and fileOrData is the name of a file, only models defined in lines are compared.

## Value

A data frame of the selected models for the proposed penalized criteria.

Wilson Toussile

## References

backward.explorer, dimJump.R.
 1 2 3 4 5 data(genotype2_ExploredModels) outDimJump = dimJump.R(genotype2_ExploredModels, N = 1000, h = 5, header = TRUE) cte1 = outDimJump[[1]][1] outSlection = model.selection.R(genotype2_ExploredModels, cte = cte1, header = TRUE) outSlection