A data scientist builds a PROC HPFOREST model and a PROC TREEBOOST model on the same data and compares their variable importance outputs. A colleague claims both should rank variables identically. Why is this assumption incorrect?

B is correct because the two procedures implement fundamentally different ensemble algorithms. PROC HPFOREST (random forest) grows trees in parallel on bootstrap samples; importance is the average Gini reduction across all trees. PROC TREEBOOST (gradient boosting) grows trees sequentially on pseudo-residuals; importance reflects how often and how much each variable reduces the loss across the boosting iterations. Different training objectives and tree dependencies produce different variable importance rankings. A is wrong; the difference is structural, not just random seed variation. C is wrong; PROC TREEBOOST does produce variable importance. D is wrong; both handle continuous and categorical variables.