While the developers have attempted to make CART easy to use and have built in a large number of convenience features, as with any statistical method, problems can arise. Help is available for some of the major problems that users may encounter in CART analyses:
- COMBINE (Bagging, ARCing) Command
- No Tree Built
- Insufficient Memory
- High Costs, or All Relative Costs Exceed One
- Cross-Validation Breakdown
- Costs are Too High
- Cannot Apply Tree to New Data
- Too Many Levels in a Categorical Predictor
- The Tree is Too Large or Complex
- Limitations on the Learn Sample Size When Using Cross Validation
- Interactive Splitting
- Internal Error Messages
- CART and Large Datasets
- Why does the tree change when non-splitting variables are dropped?
- Improvement Penalties
- Variable importance
- Nominal (ordered) Variables and Rank
- Cross Validation
- The Systat Dataset Format