The colored marbles question I know is very simple and is not statistical. "You have a jar with three colors of marbles. You can't see the marbles until you take them out of the jar. How many do you have to pull out of the jar to guarantee you have at least two of the same color?"

The question is to help me understand if you know how to look at the worst case scenario. There are three colors of marbles. The worst case is that you pick one of each color as the first 3, therefore the 4th must be the same as one of the first 3. I've gotten answers ranging from 2 to 27 to this question, and some who said it can't be solved because you don't know how many total marbles there are.

I'm not certain this is the question you're referring to, but if it is and you're approaching it as a statistical problem rather than logical, that's exactly the problem I'm trying to uncover by asking the question.

YMMV -- other interviewers may actually care about the answer and not how you got there (which I think is dumb), or you may be thinking of a different colored marbles in a jar question -- but the above is my experience on both sides of the table. Also, I sure hope this isn't the only question they ask in the interview. I have a whole list of questions that test various thought processes and for most of them, it's not the answer that matters, it's how you approached the problem and how easily you gave up (or not) that matters.