Efficient calculation of Kendall's τ for interval data

Abstract

Calculation of the strength of dependence in the case of interval data is computation-wise a very demanding task. We consider the case of Kendall's τ statistic, and calculate approximations of its minimal and maximal values using very easy to compute heuristic approximations. Using Monte Carlo simulations and more accurate calculations based on an evolutionary algorithm we have evaluated the effectiveness of proposed heuristics. © 2013 Springer-Verlag.