Partial Plots for VarPro

Plot partial effects of x-variable(s) from a VarPro analysis.

plot.partialpro(x, xvar.names, nvar,
  parametric = FALSE, se = TRUE,
  causal = FALSE, subset = NULL, plot.it = TRUE, ...)

Arguments

x

partialpro object obtained from previous call to partialpro.

xvar.names

Names of the x-variables to be plotted (uses all by default). Can also be integer values.

nvar

Number of variables to be plotted. Default is all.

parametric

Set this to true only if you think the partial effect is a polynomial.

se

Display standard errors?

causal

Display causal estimator?

subset

Optional factor for conditioning the partial effects. Does not apply if parametric='TRUE'.

plot.it

If FALSE no plots are generated and instead partial effects values are returned.

...

Additional options passed to plot.

Details

Generates a smooth partial plot for continuous variables. Smoothed black lines are the partial effect and dashed red lines indicate a smoothed error bar of +/- two standard errors. Note these "standard errors" are meant only to be a guide and should be interpreted with caution. For binary variables results are displayed as boxplots with whiskers used to indicate the standard error.

Partial effects are estimated nonparametrically using locally fit polynomial models. These are the values displayed by default and should be used in any setting where the partial effect is expected to be nonlinear. Use option parametric if the true effect is believed to be a polynomial.

A causal estimator can be requested in which case the value displayed is the baseline subtracted parametric local estimator.

Plots can be conditioned using option subset which takes the form of a factor of length equal to the original data sample size. This produces partial effects conditioned on the levels of the provided factor. subset can also be specified by a vector of logical values or a vector of integer values indicating cases of interest. In this scenario, the analysis is confined to the requested subset of the data.

Author

Min Lu and Hemant Ishwaran

References

Ishwaran H. (2025). Multivariate Statistics: Classical Foundations and Modern Machine Learning, CRC (Chapman and Hall), in press.

See also

partialpro

Examples

# \donttest{
##------------------------------------------------------------------
##
## Boston housing
##
##------------------------------------------------------------------

library(mlbench)
data(BostonHousing)
o.boston <- varpro(medv~.,BostonHousing)
oo.boston <- partialpro(o.boston, nvar=4, learner=rf.learner(o.boston))

par(mfrow=c(2,4))

## parametric local estimation (default)
plot(oo.boston, ylab="parametric est.")

## non-parametric local estimation
plot(oo.boston, parametric=FALSE, ylab="non-parametric est.")


##------------------------------------------------------------------
##
## Boston housing with subsetting
##
##------------------------------------------------------------------

library(mlbench)
data(BostonHousing)
o.boston <- varpro(medv~.,BostonHousing)
oo.boston <- partialpro(o.boston, nvar=3, learner=rf.learner(o.boston))

## subset analysis
price <- BostonHousing$medv
pricef <- factor(price>median(price), labels=c("low priced","high priced"))
par(mfrow=c(1,1))
plot(oo.boston, subset=pricef, nvar=1)


##------------------------------------------------------------------
##
## veteran data with subsetting using celltype as a factor
##
##------------------------------------------------------------------

data(veteran, package = "randomForestSRC")
dta <- veteran
dta$celltype <- factor(dta$celltype)
o.vet <- varpro(Surv(time, status)~., dta)
oo.vet <- partialpro(o.vet, nvar=6, nsmp=Inf, learner=rf.learner(o.vet))

## partial effects, with subsetting
par(mfrow=c(2,3))
plot(oo.vet, subset=dta$celltype)

## causal effects, with subsetting
par(mfrow=c(2,3))
plot(oo.vet, subset=dta$celltype, causal=TRUE)

# }