plot.quantreg.rfsrc: Plot Conditional Quantiles and CRPS Diagnostics

In randomForestSRC: Fast Unified Random Forests for Survival, Regression, and Classification (RF-SRC)

Plot Conditional Quantiles and CRPS Diagnostics

Description

Plots estimated conditional response quantiles against observed responses. An optional inset compares the forest's continuous ranked probability score (CRPS) with the score for the empirical CDF of training responses. Mean pinball losses can also be displayed for selected quantiles.

Usage

## S3 method for class 'rfsrc'
plot.quantreg(x, prbL = .25, prbU = .75,
  m.target = NULL, crps = TRUE, subset = NULL,
  xlab = NULL, ylab = NULL, ...,
  inset.args = list(), crps.null = TRUE, quantreg.tau = NULL)

Arguments

x	A training or prediction object returned by quantreg, with observed responses for the observations to be plotted. For new-data diagnostics, include response columns in the quantreg() prediction call.
prbL, prbU	Lower and upper quantile levels: single probabilities in (0,1) with prbL < prbU. The middle level is 0.5 if it lies strictly between them, or their midpoint otherwise. get.quantile() matches each level to the nearest stored probability. Include the desired levels in prob when calling quantreg() for exact matching.
m.target	Name of the continuous response to display. Defaults to the first response with quantile results. Selects existing results without calculating new predictions.
crps	If TRUE, show the standardized CRPS curve from get.quantile.crps() in an inset. Scores use the finite CDF reporting grid.
subset	Rows to plot and score. Supply positive integer indices or a logical vector with one nonmissing entry per row of the quantile output. NULL uses all rows. Indices refer to results after training or prediction preprocessing; repeated indices count a row more than once.
xlab, ylab	Main-panel axis labels. Defaults are the selected response name and "Target Quantiles".
...	Named graphical arguments passed to plot.default for the main panel, including xlim, ylim, main, cex.axis, and las. Use pch, col, and cex to style the middle-quantile points. The function sets the coordinates and initial panel type. These arguments do not style the inset.
inset.args	Named list of graphical arguments passed to plot.default for the CRPS inset. Use xlim and ylim for limits; col, lty, and lwd for the forest curve; and cex.axis, las, tck, or related options for the axes. axes = FALSE hides both axes; yaxt = "n" hides the right-hand axis. The default list() uses the inset's own settings, with limits covering both curves. Ignored when crps = FALSE. Supply the full argument name.
quantreg.tau	Quantile levels for mean pinball-loss labels, as a vector of finite probabilities strictly between zero and one. Losses use the selected response and rows and appear in the lower-right corner of the main panel. NULL omits labels even when the session option quantreg.tau is set. The plotted quantiles and CRPS inset are unchanged. Supply the full argument name.
crps.null	If TRUE, add a dashed black null curve and a Forest/Null legend to the inset. FALSE shows the forest curve alone. Training responses for the null distribution come from x$forest$yvar, or from x$yvar for a training object without responses saved in its forest. If unavailable, omit the null curve with a warning. Ignored when crps = FALSE. Supply the full argument name.

Details

Each observation is plotted at its observed response on the horizontal axis. A point marks its predicted middle quantile, and a vertical segment with endpoint marks spans its lower and upper quantiles. The segment is a prediction interval for an individual response, not a confidence interval for the conditional mean. The dashed diagonal marks equality between observed and predicted values. Defaults request the median and the 25th and 75th percentiles. The response selected by m.target and rows selected by subset apply to the panel and all scores.

Horizontal jitter separates overlapping observations; scores use the original response values. Rows with an unavailable response or plotted quantile are omitted from the panel with a warning.

To add pinball losses, supply levels such as quantreg.tau = c(.1, .5, .9). They can differ from the levels drawn in the panel. Losses are calculated by get.pinball.error(x, tau = quantreg.tau, subset = subset, m.target = m.target). Each mean uses selected rows with finite responses and quantile predictions at that level. Smaller losses indicate better quantile predictions.

The inset shows the standardized CRPS curve from get.quantile.crps(). At each response threshold on the horizontal axis, the right-hand axis shows prediction error integrated from the first grid value to that threshold and divided by the integration width. Smaller values indicate less error. Scores depend on the grid's range and resolution; see quantreg for the calculation. The first value is unavailable because its integration width is zero. An entirely unavailable curve is omitted with a warning.

The null distribution is the empirical CDF of all finite training responses for the selected outcome and is unchanged by subset. Its CRPS curve uses the same grid, observations, integration, and standardization as the forest curve. At each threshold, both scores exclude rows with an unavailable response or forest CDF prediction. For training plots, the null distribution includes the evaluated responses and is an in-sample reference, even for OOB predictions.

Use ... to customize the main panel and inset.args for the inset. Axis limits change only the displayed region, not the rows scored or the CRPS integration limits. After drawing the inset, graphics settings return to the main panel so that further drawing applies there. Successive calls can be used in a multi-panel layout.

Value

Used for its graphical side effect. Invisibly returns NULL.

Author(s)

Hemant Ishwaran and Udaya B. Kogalur

See Also

quantreg, get.quantile, get.quantile.crps, get.pinball.error, plot.default

Examples

## Univariate quantiles, with forest and null CRPS curves.
library(randomForestSRC)
set.seed(19)
dta <- na.omit(airquality)
prob <- c(.25, .50, .75)
q <- quantreg(Temp ~ ., data = dta, prob = prob, ntree = 100)
plot.quantreg(q)

## Add pinball losses for this plot only; the interval and inset are unchanged.
plot.quantreg(q, quantreg.tau = c(.2, .5, .8))
print(get.pinball.error(q, tau = c(.2, .5, .8)))

## Customize the main panel directly, and the inset separately.
plot.quantreg(q, main = "Temperature",
              xlim = c(50, 100), ylim = c(45, 110), pch = 19,
              inset.args = list(xlim = c(60, 95), ylim = c(0, .35),
                                lwd = 1.5))

## Keep the inset, but suppress its null reference.
plot.quantreg(q, crps.null = FALSE)

## Multivariate forest: select each response for plotting only.
mv <- quantreg(cbind(Ozone, Temp) ~ ., data = dta,
               splitrule = "mahalanobis", prob = prob, ntree = 100)
print(names(get.quantile(mv, pretty = FALSE)))
op <- par(mfrow = c(1, 2))
plot.quantreg(mv, m.target = "Ozone")
plot.quantreg(mv, m.target = "Temp", main = "Temperature",
              xlim = c(50, 100), ylim = c(45, 110),
              inset.args = list(ylim = c(0, .35)))
par(op)

randomForestSRC documentation built on Nov. 1, 2026, 5:07 p.m.