The HPFOREST Procedure

If the value of a target variable is missing, the observation is excluded from training and from evaluating the model. If the value of an input variable is missing, PROC HPFOREST uses the missing value as a legitimate value by default or if MISSING=USEINSEARCH and the number of observations in which the splitting variable has missing values is at least as large as the value of the MINUSEINSEARCH= option.

The discussion in this section applies to each candidate variable and each node separately. For example, the test of association that uses input variable X might use all observations, and the test that uses input variable Z might ignore some observations because of missing values. The test that uses X might use all observations in one node but not all observations in another node.