plot.registab: Diagnostic plotting to see if RCP groups are stable

In RCPmod: Regions of Common Profiles Modelling with Mixtures-of-Experts

Diagnostic plotting to see if RCP groups are stable

Description

For increasing size of hold-out samples, cooks distance and predictive log-likelihood are plotted.

Arguments

You can put normal text in Arguments, too, like this. Remember to indent all arguments, as below.

x	a registab object (the output of stability.regimix)
y	ignored.
minWidth	the minimum width of the density for each hold-out size. A rectangle of minWidth (at least) will be placed, and centered, at each of the hold-out sizes.
ncuts	number of cuts to use to describe the distributions of predicted log-likelihood. This number is a maximum value and the actual number used will depend on the data (see the breaks argument of hist)
ylimmo	y-limits on the predicted log-likelihood plot. Typically, upper bound will be zero. If NULL (default), the lower bound is taken to be the smallest observation in the data.
...	ignored

Value

Nothing, but causes plots to be produced on the graphics device

Method

plot.registab(x, y, minWidth=1, ncuts=111, ylimmo, ...)

See Also

stability.regimix, regimix, cooks.distance.regimix

Examples

## Not run: 
#not run as R CMD check complains about the time taken.
#This code will take a little while to run (about 3.5minutes on my computer)
system.time({
example( regimix);
my.registab <- stability.regimix( fm, oosSizeRange=seq( from=1,to=fm$n%/%5,length=5),
      times=fm$n, mc.cores=2, doPlot=FALSE);
plot( my.registab, minWidth=1, ncuts=15);
})

## End(Not run)

RCPmod documentation built on Oct. 26, 2022, 1:09 a.m.