[R] TR: Latent Class Analysis

[Robert Mcfadden]

> De : Pousset [mailto:maud.pousset w noos.fr]
> Envoyi : mardi 4 juillet 2006 18:38
> @ : 'r-help w stat.math.ethz.ch'
> Objet : Latent Class Analysis
> 
> 
> 
> Hello everybody,
> 
> 
> 
> I am working on latent class analysis and have already used the R
> function
> + lca ; (in the e1071 package). I ve got interesting results but I cant
> simply find out the methodology used by this routine :
> 
> 1) What kind of model is behind the routine (mixture model? If so, can you
> choose among different kind of distributions such as normal, Poisson,
> binomial)
> 
> 2) What kind of algorithm is used (hierarchical methods? Relocation
> methods?)
> 
> 3) Which criterion allows determining the best model?
> 
> In addition, I wonder if it is possible, with R software, to determine the
> best number of class or do one have to fix it a priori.
> 
> If one can help, thanks a lot,
> 
> 
> 
> Maud
> 
> INSERM U669, Cochin Hospital, Paris
> 
> 
> 


You can use package (available at R page) poLCA that has documentation
describing what you want. Look also at http://dlinzer.bol.ucla.edu/poLCA/
As far as you know BIC, AIC (based on chi-sqr and G^2 statistics) and
Cressie-Read allow you to choose the most appropriate model.
Best,
Robert