Visualise probabilistic regression distribution predictions

Plots probability density functions from n predicted probability distributions.

Usage

plot_probregr(
  p,
  n,
  type = c("point", "line", "both", "none"),
  which_plot = c("random", "top"),
  rm_zero = TRUE,
  ...
)

Arguments

p

(PredictionRegr)
With at least column distr.

n

(integer(1))
Number of predictions to plot.

type

(character(1))
One of "point" (default), "line", "both", "none".

which_plot

(character(1))
One of "random" (default) or "top". See details.

rm_zero

(logical(1))
If TRUE (default) does not plot points where f(x) = 0.

...

Unused

Details

type:

• "point" (default) - Truth plotted as point (truth, predicted_pdf(truth))

• "line" - Truth plotted as vertical line intercepting x-axis at the truth.

• "both" - Plots both the above options.

• "none" - Truth not plotted (default if p$truth is missing).

which_plot:

• "random"(default) - Random selection ofn` distributions are plotted.

• "top"- Topn` distributions are plotted.

It is unlikely the plot will be interpretable when n >> 5.

Examples

if (FALSE) {
library(mlr3verse)
task = tsk("boston_housing")
pipe = as_learner(ppl("probregr", lrn("regr.ranger"), dist = "Normal"))
p = pipe$train(task)$predict(task)
plot_probregr(p, 10, "point", "top")
}