An alternative approach to dimension reduction for pareto distributed data: a case study