Barplot in palce of point pdp plot for categorical variables using autoplot function

[bappa10085]

I think it is better to use barplot in palce of point pdp plot for categorical variables using autoplot function as suggested here. I am rpoviding a minimal reproduicible example here

library(pdp)

# dummy data
categorical <- c('A', 'A', 'A', 'A', 'B', 'B', 'B', 'B')
numerical <- c(1, 2, 3, 4, 1, 2, 3, 4)
target <- c(100, 200, 300, 400, 500, 600, 700, 800)
data <- data.frame(categorical, numerical, target)
data$categorical <- factor(data$categorical)

set.seed(101)  # for reproducibility
mod.rf <- randomForest(target ~ ., data = data)

cat.pdp <- partial(mod.rf, pred.var = c("categorical"))

autoplot(cat.pdp, contour = TRUE) 

Then I have tried the following code to generate barplot

autoplot(cat.pdp, contour = TRUE) +
  geom_col()

As we can see, the points are also plotted and the y-axis limit is starting from 0 unlike the point plot where y-axis limits are optimised according to the data.

It would be better if the point plot is repalced by barplot.

[bgreenwell]

Thanks @bappa10085, I’ll plan to add geom and aes arguments to make it more flexible.