Skip to contents

Create a 2-way partial dependence plot for deweather models

Source: R/plot_dw_partial_2d.R
plot_dw_partial_2d.Rd

Generates 2-way partial dependence plot to visualize the relationship between two predictor variables and model predictions. These plots show how the predicted pollutant concentration changes as a function of two variables while averaging over the effects of all other variables.

Usage

plot_dw_partial_2d(
  dw,
  var_x = NULL,
  var_y = NULL,
  intervals = 40L,
  contour = c("none", "lines", "fill"),
  contour_bins = 8,
  exclude_distance = 0.05,
  show_conf_int = FALSE,
  n = NULL,
  prop = 0.01,
  cols = "viridis",
  radial_wd = FALSE,
  ...,
  .plot = TRUE,
  .plot_engine = NULL,
  .progress = rlang::is_interactive()
)

Arguments

dw

A deweather model created with build_dw_model().

var_x, var_y

The name of the two variables to plot. Must be one of the variables used in the model. If both are missing, the top two most individually important numeric variables will be selected automatically.

intervals

The number of points for the partial dependence profile.

contour

Show contour lines on the plot? Can be one of "none" (the default, no contour lines), "lines" (draws lines) or "fill" (draws filled contours using a binned colour scale).

contour_bins

How many bins should be drawn if contour != "none"?

exclude_distance

A 2-way partial dependence plot uses mgcv::exclude.too.far() to ensure the plotted surface is within range of the original input data. exclude_distance defines how far away from the original data is too far to plot. This should be in the range 0 to 1, where higher values are more permissive; 1 will retain all data.

show_conf_int

Should the bootstrapped 95% confidence interval be shown? In plot_dw_partial_2d() this creates separate facets for the lower and higher confidence intervals. It may be easiest to see the difference by using contour = "fill".

n

The number of observations to use for calculating the partial dependence profile. If NULL (default), uses prop to determine the sample size.

prop

The proportion of input data to use for calculating the partial dependence profile, between 0 and 1. Default is 0.01 (1% of input data). Ignored if n is specified.

cols

Colours to use for plotting. See openair::openColours().

radial_wd

Should the "wd" (wind direction) variable be plotted on a radial axis? This can enhance interpretability, but makes it inconsistent with other variables which are plotted on cartesian coordinates. Defaults to FALSE.

...

Not currently used.

.plot

When FALSE, return a data.frame of plot data instead of a plot.

.plot_engine

The plotting engine to use. One of "ggplot2", which returns a static plot, or "plotly", which returns a dynamic HTML plot.

.progress

Show a progress bar? Defaults to TRUE in interactive sessions.

Value

A ggplot2 object showing the partial dependence plot. If plot = FALSE, a named list of plot data will be returned instead.

Parallel Processing

This function supports parallel processing using the {mirai} package. You will likely find that the performance of this function increases if "daemons" are set. The greatest benefits will be seen if you spawn as many daemons as you have cores on your machine, although one fewer than the available cores is often a good rule of thumb.

# set workers - once per session
mirai::daemons(4)

# run your function as normal
tune_dw_model(aqroadside, "no2", tree_depth = c(5, 10))

Plotting Engines

deweather offers different plotting engines for different purposes. At the moment, two plotting engines are supported:

  • "ggplot2", for static plotting. This engine produces plots which can be easily saved to a .png, .svg, or other 'static' file format. To save a ggplot2 plot, it is recommended to use the ggplot2::ggsave() function.

  • "plotly", for dynamic plotting. This engine produces HTML plots which are suitable for embedding into quarto or rmarkdown documents, or for use in shiny applications. These can be saved using functions like htmlwidgets::saveWidget().

The plotting engine is defined using the .plot_engine argument in any plot_* function in deweather.

# use default
plot_dw_importance(dw)

# set to ggplot2 (static)
plot_dw_importance(dw, .plot_engine = "ggplot2")

# set to plotly (HTML)
plot_dw_importance(dw, .plot_engine = "plotly")

When .plot_engine is not set, the engine defaults to "ggplot2". However, this option can be overridden by using the deweather.plot_engine global option.

# set once per session
options("deweather.plot_engine" = "plotly")

# now defaults to "plotly" over "ggplot2"
plot_dw_importance(dw)

Note that not all arguments in a function may apply to all plotting engines. For example, hexagonal binning in plot_tdw_testing_scatter() is supported in ggplot2 but not in plotly at time of writing.

Author

Jack Davison

Examples

if (FALSE) { # \dontrun{
dw <- build_dw_model(aqroadside, "no2")
plot_dw_partial_2d(dw)
} # }