There is no inherent limit on the size of the decision tree model that can be evaluated and analyzed with the DTREE procedure. The numbers of stages and outcomes are constrained only by the amount of memory available. Naturally, there needs to be a sufficient amount of core memory available in order to invoke and initialize the SAS system. Furthermore, more memory is required to load the graphics sublibrary if the GRAPHICS option is specified. As far as possible, the procedure attempts to store all the data in core memory. However, if the problem is too large to fit in core memory, the procedure resorts to the use of utility data sets and swaps between core memory and utility data sets as necessary.
The storage requirement for the data area required by the procedure is proportional to the number of stages and outcomes as well as the number of nodes[2] in the decision tree model. The time required depends heavily on the number of nodes in the decision tree.