ggRandomForests (version 3.1.2)

plot.gg_variable: Plot a gg_variable object,

Description

Plot a gg_variable object,

Usage

# S3 method for gg_variable
plot(
  x,
  xvar,
  time,
  time_labels,
  panel = FALSE,
  oob = TRUE,
  points = TRUE,
  smooth = TRUE,
  ...
)

Value

A single ggplot object when length(xvar) == 1 or panel = TRUE; otherwise a patchwork composite that stacks one panel per variable in xvar. Either way the result is one plottable object, never a bare list, so it composes with patchwork and dispatches through ggplot2::autoplot(). For the patchwork case, to inspect one panel with ggplot2::layer_data() pull that panel out first (e.g. ggplot2::layer_data(p[[1]])).

Arguments

x

gg_variable object created from a rfsrc object

xvar

variable (or list of variables) of interest.

time

For survival, one or more times of interest

time_labels

string labels for times

panel

Should plots be faceted along multiple xvar?

oob

oob estimates (boolean)

points

plot the raw data points (boolean)

smooth

include a smooth curve (boolean)

...

arguments passed to the ggplot2 functions.

References

Breiman L. (2001). Random forests, Machine Learning, 45:5-32.

Ishwaran H. and Kogalur U.B. (2007). Random survival forests for R, Rnews, 7(2):25-31.

Ishwaran H. and Kogalur U.B. randomForestSRC: Random Forests for Survival, Regression and Classification. R package version >= 3.4.0. https://cran.r-project.org/package=randomForestSRC

See Also

gg_variable, gg_partial, plot.variable

Examples

## ------------------------------------------------------------
## classification
## ------------------------------------------------------------
## -------- iris data
set.seed(42)
rfsrc_iris <- randomForestSRC::rfsrc(Species ~ ., data = iris, ntree = 50)

gg_dta <- gg_variable(rfsrc_iris)
plot(gg_dta, xvar = "Sepal.Width")
plot(gg_dta, xvar = "Sepal.Length")

## Panel plot across all predictors
plot(gg_dta,
  xvar = rfsrc_iris$xvar.names,
  panel = TRUE, se = FALSE
)

## ------------------------------------------------------------
## regression
## ------------------------------------------------------------
## -------- air quality data
# na.action = "na.impute" handles missing Ozone / Solar.R values
set.seed(42)
rfsrc_airq <- randomForestSRC::rfsrc(Ozone ~ ., data = airquality,
                    na.action = "na.impute", ntree = 50)
gg_dta <- gg_variable(rfsrc_airq)

# Treat Month as an ordinal factor for better visualisation
gg_dta[, "Month"] <- factor(gg_dta[, "Month"])

plot(gg_dta, xvar = "Wind")
plot(gg_dta, xvar = "Temp")
plot(gg_dta, xvar = "Solar.R")

# Factor variable uses notched boxplots
plot(gg_dta, xvar = "Month", notch = TRUE)

# \donttest{
# Panel plot across continuous predictors (loess smooths; slower)
plot(gg_dta, xvar = c("Solar.R", "Wind", "Temp", "Day"), panel = TRUE)
# }

## ------------------------------------------------------------
## survival examples
## ------------------------------------------------------------
## -------- veteran data
# \donttest{
data(veteran, package = "randomForestSRC")
set.seed(42)
rfsrc_veteran <- randomForestSRC::rfsrc(Surv(time, status) ~ ., veteran,
  nsplit = 10,
  ntree = 50
)

# Marginal survival at 90 days
gg_dta <- gg_variable(rfsrc_veteran, time = 90)

# Single-variable dependence plots
plot(gg_dta, xvar = "age")
plot(gg_dta, xvar = "diagtime")

# Panel coplot for two predictors at a single time
plot(gg_dta, xvar = c("age", "diagtime"), panel = TRUE)

# Compare survival at 30, 90, and 365 days simultaneously
gg_dta <- gg_variable(rfsrc_veteran, time = c(30, 90, 365))

# Single-variable plot (one facet per time point)
plot(gg_dta, xvar = "age")

# Panel coplot across two predictors and three time points
plot(gg_dta, xvar = c("age", "diagtime"), panel = TRUE)
# }