Frank: I do not understand your answer. For me it is just a bug of the program. Just imagine the following situation:

Instead of growing the tree automatically, select the same predictor than the algorithm would have choosen automatically (select the best predictor). It happened to me to see final nodes with less than 50 cases, when I had fixed for growing criteria : minimun number of cases in child node:50. and growing method C&RT, I think is wrong. If not, I will thank you for explaining to me why not.

Author: "Dr. Frank Knapp" <frank.knapp@psyma.com> Date: 12/04/99 3:43 PM

(just the last part of the original message:)

And to the problem of Maria Elena:

If you force a split by a specific variable, the constraint (number of cases) won't be considered (otherwise, you wouldn't be able to look at the results). Or did I misunderstand your problem?