Random Forest partial plot for a binary predictor

Question

I'm not sure how to interpret a partial dependence plot for a categorical variable. For instance, the partial dependence (command partialPlot from the randomForest package) for a binary predictor would give two values, one for each category.

How exactly do you interpret those values? The documentation of partialPlot is quite cryptic in this respect.

My confusion arises, I guess because I'm thinking in terms of a logistic regression, where with a dummy coding system you in general obtain the log-odds of the variable of interest against the baseline category.

Answer 1

In general partial plots are the average of a series of hypothetical predictions.

For continuous variables, partialPlot selects a series of n.pt values along the variable of interest. It creates a test data set identical to the training data, and sequentially sets the variable of interest for all observations to a selected value. It then takes the average value of the predicted response for each test set and plots the results against the select value. For a binary categorical predictor, this process would result in only two values.

It is helpful to use partial plots in combination with variable dependence and interpret them together. The variable dependence figure is generated by plotting the forest predicted value against the variable of interest for each observation.