Remove original data from regression_forest without changing the type to "list"-CodePudding

I would like to estimate a regression_forest with the grf package and remove the original data that is stored in the regression_forest output for data protection reasons.

The problem is that when I remove the data, R doesn't recognize the object as a regression_forest anymore and therefore throws an error.

Does anyone know how to go around this problem?

Here is a reproducible example:

library(grf)

# Train a standard regression forest.
n <- 50
p <- 10
X <- matrix(rnorm(n * p), n, p)
Y <- X[, 1] * rnorm(n)
r.forest <- regression_forest(X, Y)

# Remove the original data
r.forest <- r.forest[-c(18,19)]

# Predict using the forest.
X.test <- matrix(0, 101, p)
X.test[, 1] <- seq(-2, 2, length.out = 101)
r.pred <- predict(r.forest, X.test)

The last line causes the following error:

Error in UseMethod("predict") : no applicable method for 'predict' applied to an object of class "list"

CodePudding user response：

The predict function seems to need to know the dimensions of the original data, but from what I can tell it doesn't need the data itself.

If you convert the original data stored in the model object to NA, then the predictions seem unaffected.

# Get original predictions
r.pred.original <- predict(r.forest, X.test)

# Convert stored data to NA
r.forest$X.orig[!is.na(r.forest$X.orig)] <- NA

# Get new predictions
r.pred.new <- predict(r.forest, X.test)

# r.pred.original and r.pred.new are the same