The aim of this paper is to extract mathematically talented students out of a group of arbitrary high school students. We do this by applying a stepwise discriminant analysis modified for ordinal data to the results of German high school students at the international mathematics competition ?Kangaroo of Mathematics?. It turns out that three of the thirty given problems are enough to discriminate between laureates, which we assume to be mathematically talented, and non-laureates. The three chosen problems are from different mathematical fields.
Discriminant analysis for ordinal data discrete kernel estimation testing mathematical talent multiple choice competition