pdp_ice_ale: Plot PD, ICE, and ALE curves

In mirka-henninger/InterpretationMethods: Facilitates the implementation of interpretation techniques for social science research

Description

This function allows to create PD, ICE, ALE plots ICE plots can be centered at the minimum Plots for several features of one model can be created at the same time

Usage

pdp_ice_ale(
  pred,
  features,
  method,
  xlabel = features,
  ylabel = method,
  title = "",
  center = FALSE,
  limits = c(NA, NA),
  nCol = NA,
  alpha = 0.5,
  sample_prop = 0.5
)

Arguments

pred	A prediction object from package iml
features	A character vector containing the names of the features for which the plot should be created
method	A character string indicating the method to be applied: either "pdp", "pdp+ice", "ice", or "ale"
xlabel	An optional character vector indicating x-axis label
ylabel	An optional character string indicating y-axis label
title	An optional character string indicating the title of the plot
center	Logical indicating whether ICE curves should be centered at the minimum. Default is FALSE
limits	An optional two-entry vector indicating the limits of the y-axis
nCol	An optional numeric entry indicating the number of columns when plots are created for several features
alpha	An optional numeric entry indicating the alpha-level of ICE curves
sample_prop	An optional numeric entry between 0 and 1 indicating the percentage of individiuals that should be sampled for ICE curves

Value

a plot of type ggplotify

Examples

## Not run: 
N <- 1000
x1 <- runif(N, -1, 1)
x2 <- runif(N, -1, 1)
y <- 5 + 2 * x1 + x2 + rnorm(N,1)
dat <- data.frame(x1,x2,y)
rfmod <- randomForest::randomForest(y~., dat)
pred <- iml::Predictor$new(rfmod)
pdp <- pdp_ice_ale(pred,  "x1", method = "pdp")
ice <- pdp_ice_ale(pred,  "x1", method = "ice", alpha = .1)
pdp_ice <- pdp_ice_ale(pred,  "x1", method = "pdp+ice", alpha = .2)
ale <- pdp_ice_ale(pred,  c("x1", "x2"), method = "ale", ylabel = "ALE")

## End(Not run)

mirka-henninger/InterpretationMethods documentation built on Jan. 12, 2022, 4:10 p.m.