summaryplot: Summary of models

Description

This function provides a plot summarizing the results of different models fitted by hlme, lcmm, multlcmm, Jointlcmm, mpjlcmm or externVar.

Usage

summaryplot(
  m1,
  ...,
  which = c("BIC", "entropy", "ICL"),
  mfrow = c(1, length(which)),
  xaxis = "G"
)

Arguments

m1: an object of class hlme, lcmm, multlcmm, Jointlcmm, mpjlcmm, externVar or externVar.
...: further arguments, in particular other objects of class hlme, lcmm, multlcmm, Jointlcmm or mpjlcmm, and graphical parameters.
which: character vector indicating which results should be plotted. Possible values are "loglik", "conv", "npm", "AIC", "BIC", "SABIC", "entropy", "ICL", "ICL1", "ICL2".
mfrow: for multiple plots, number of rows and columns to split the graphical device. Default to one line and length(which) columns.
xaxis: the abscissa of the plot. Default to "G", the number of latent classes.

Author

Sasha Cuau, Viviane Philipps, Cecile Proust-Lima

Details

Can be reported the usual criteria used to assess the fit and the clustering of the data: - maximum log-likelihood L (the higher the better) - number of parameters P, number of classes G, convergence criterion (1 = converged) - AIC (the lower the better) computed as -2L+2P - BIC (the lower the better) computed as -2L+ P log(N) where N is the number of subjects - SABIC (the lower the better) computed as -2L+ P log((N+2)/24) - Entropy (the closer to one the better) computed as 1+sum[pi_ig*log(pi_ig)]/(N*log(G)) where pi_ig is the posterior probability that subject i belongs to class g - ICL (the lower the better) computed in two ways : ICL1 = BIC - sum[pi_ig*log(pi_ig)] or ICL2 = BIC - 2*sum(log(max(pi_ig)), where the max is taken over the classes for each subject. - %class computed as the proportion of each class based on c_ig

Examples

Run this code

if (FALSE) {
library(NormPsy)
paquid$normMMSE <- normMMSE(paquid$MMSE)
paquid$age65 <- (paquid$age - 65)/10
m1 <- hlme(normMMSE~age65+I(age65^2)+CEP, random=~age65+I(age65^2), subject='ID', data=paquid)
m2 <- hlme(normMMSE~age65+I(age65^2)+CEP, random=~age65+I(age65^2), subject='ID', data=paquid,
ng = 2, mixture=~age65+I(age65^2), B=m1)
m3g <- gridsearch(hlme(normMMSE~age65+I(age65^2)+CEP, random=~age65+I(age65^2), subject='ID',
data=paquid, ng=3, mixture=~age65+I(age65^2)), rep=100, maxiter=30, minit=m1)
summaryplot(m1, m2, m3g, which=c("BIC","entropy","ICL"),bty="l",pch=20,col=2)
}