Wondering if there are any tricks to modeling decision trees in an RDBMS? What I've got is a slate of questions that are answered as True/False. The next question to be asked is dependent on the answer to the previous question. At the end, I get a conclusion based on the series of answers. The number of questions necessary to derive a conclusion is variable depending on the branching.

For the sake of argument, here's a simple example:

Did you vote in the last election?\n   Yes:\n      Did you vote for Bush?\n         Yes: You are an idiot\n         No:  You are a fool\n   No:\n      Are you under 18?\n         Yes: Get ready to be drafted\n         No:  Are you stupid?\n            Yes: You are stupid\n            No:  You are dumb

No shared questions (at least for the time being).

There is a convergence possible, but only in the conclusions. That is, you don't have two or more paths to the same question.

If you don't count NULL, then only two answers are possible in the current scheme of things.

(As a slight background, these are questions related to accounting control deficiencies as applied to Sarbanes-Oxley).