boxplot.choicemod: Parrallel boxplots of choicemod

In chavent/modvarsel: Implements a computational methodology for model and variables selection

Description

The parallel boxplots are built with the MSE values obtained at each replication of the procedure implemented in choicemod. At each replication, each regression method estimates both the reduced model (with covariates selection) and the complete models (without covariates selection). The boxplots represent the test mean square errors (performed on test samples) of each regression method including (or not) covariates selection.

Usage

## S3 method for class 'choicemod'
boxplot(res, method = NULL, type = "reduced",
  ylab = NULL, ...)

Arguments

res	an object of class choicemod.
method	if NULL, all the methods in res$method are plotted. Otherwise a subset of these methods ("linreg", "sir", "rf", "pcr", "plsr", "ridge").
type	this should be "reduced" (boxplot for reduced estimated models), "complete" (boxplot for complete estimated models) or "both". By default, type="reduced".
ylab	Y axis label.
...	further arguments passed to or from other methods.

See Also

choicemod, barplot.choicemod

Examples

data("simus")
res <- simus$res1
class(res)
res$method
boxplot(res, type="reduced", col="gold", main="Models with variables selection, N=10")
boxplot(res,type="complete",col=rgb(0,0,1,0.2), main="Models with all the variables, N=10")
boxplot(res,type="both", col = "aquamarine", las = 2, main = "N = 10 replications")

chavent/modvarsel documentation built on May 22, 2019, 2:22 p.m.