{smcl} {* *! version 1.0.0 10June2011, Joerg Luedicke}{...} {hline} help {hi:fmmlc} {hline} {title:Postestimation for finite mixture models} {p 8 17 2}{cmd:fmmlc} [ {cmd:,} {cmdab:savec} {cmdab:savep} ] {title:Description} {p 4 4 2} {cmd:fmmlc} is a postestimation command for {cmd:fmm}. {cmd:fmm} is a user-written command (Deb 2008) that fits finite mixture regression models. Being called after {cmd:fmm}, {cmd:fmmlc} provides a convenient overview of latent class posterior probabilities, classification of subjects based on most likely latent class membership, average posterior probabilities by latent class, the quality of a classification (entropy), and the information criteria AIC, BIC, and a sample size adjusted BIC. {title:Options} {p 4 8 2}{cmd:savec} saves the categorical latent variable based on most likely latent class membership. When using this option, a variable with the name -_class_1- will be saved. If this variable already exists, a variable with the name -_class_2- will be saved, and so on. {p 4 8 2}{cmd:savep} saves latent class posterior probabilities. When using this option, predicted posterior probabilities are saved as variables. The number of variables corresponds to the number of components from the mixture model and are indexed accordingly. {title:Technical note} {p 4 4 2} The entropy measure implemented in {cmd:fmmlc} uses formula 24 in Ramaswamy et al. (1993). Calculation of the Akaike information criterion (AIC) follows Akaike (1987). In the calculation of the Bayesian information criterion (BIC, Schwartz 1978), the number of observations from the previous {cmd:fmm} run is used (also see: {help bic_note}). The sample size adjusted BIC follows Sclove's (1987) suggestion to replace the number of observations (n) with ((n+2)/24). {title:Saved results} {pstd} {cmd:fmmlc} saves the following in {cmd:r()}: {synoptset 15 tabbed}{...} {p2col 5 15 19 2: Scalars}{p_end} {synopt:{cmd:r(AIC)}}Akaike information criterion{p_end} {synopt:{cmd:r(BIC)}}Bayesian information criterion{p_end} {synopt:{cmd:r(BIC2)}}Sample size adjusted BIC{p_end} {synopt:{cmd:r(entropy)}}Entropy measure{p_end} {synoptset 15 tabbed}{...} {p2col 5 15 19 2: Matrix}{p_end} {synopt:{cmd:r(app)}}Average posterior probabilities{p_end} {title:Examples} {p 4 8 2}{inp:. webuse nlsw88} {p 4 8 2}{inp:. gen lnwage=log(wage)} {p 4 8 2}{inp:. fmm lnwage tenure age, components(2) mix(normal)} {p 4 8 2}{inp:. fmmlc} {p 4 8 2}{inp:. fmm lnwage tenure age, components(3) mix(normal)} {p 4 8 2}{inp:. fmmlc, savec savep} {title:References} {p 4 4 2} Akaike, H. (1987). Factor analysis and AIC. {it:Psychometrika, 52}, 317–332. {p 4 4 2} Deb, P. (2008). FMM: Stata module to estimate finite mixture models. {it:Statistical Software Components, Boston College Department of Economics}, {browse "http://econpapers.repec.org/RePEc:boc:bocode:s456895"}. {p 4 4 2} Ramaswamy, V., DeSarbo, W., Reibstein, D., and Robinson, W. (1993). An empirical pooling approach for estimating marketing mix elasticities with PIMS data. {it:Marketing Science, 12}, 103–124. {p 4 4 2} Schwartz, G. (1978). Estimating the dimension of a model. {it:The Annals} {it:of Statistics, 6}, 461–464. {p 4 4 2} Sclove, L. (1987). Application of model-selection criteria to some problems in multivariate analysis. {it:Psychometrika, 52}, 333–343. {title:Author} {p 4 4 2}Joerg Luedicke, Yale University, Department of Psychology, New Haven, USA{break} {p 4 4 2}email: joerg.luedicke1@gmail.com {title:Also see} {p 4 13 2}Help {help fmm} and {help fmm postestimation} (if installed)