NA handling regression forest vs. local linear forest covariate matrices

[spocksdad]

Description of the bug
Regression forest allows the X matrix to have some incomplete cases, but local linear forest returns an error. Not sure if there's some technical reason why ll forests can't handle incomplete cases?

Steps to reproduce

#Toy data 
Y    <- as.vector(rnorm(100))
X    <- data.frame(x1 = rnorm(100), x2 = rnorm(100))
#
#Add NAs 
X$x1 <- ifelse(X$x1 > 0,X$x1,NA)
#
#Let's try an r forest 
regression_forest(Y = Y,
                  X = X)
#R forest runs fine 
#
# Now let's try ll forest 
ll_regression_forest(Y = Y,
                     X = X)
#
#ll forest returns: Error in validate_X(X) : The feature matrix X contains at least one NA.

GRF version
GRF version 2.2.0

[erikcs]

Hi @spocksdad, ll forest doesn't support NA in X because it runs OLS with X.

[erikcs]

Btw, you could stitch together your own ll forest that allows NAs in Xjs which are not in ll.split.variables by removing that input check and making sure you call the forest only with missing in features not used in ll corrections (will likely not modify grf to support this anytime soon).

[joepvdburg]

Btw, you could stitch together your own ll forest that allows NAs in Xjs which are not in ll.split.variables by removing that input check and making sure you call the forest only with missing in features not used in ll corrections (will likely not modify grf to support this anytime soon).

I tried to do this; I've removed the input check and deleted the rows data with NA values in 'll.split.variables', so that only NA values exist outside 'll.split.variables'. But my out-of-bag predictions now give a NULL value. How can this be solved? Thank you in advance

[erikcs]

I guess you'd have to call predict(ll.forest, linear.correction.variables = variables without missing values).

[joepvdburg]

Ah yes I missed that indeed. Thanks