Improvements in Computation and Usage of Joint CDFs for the N-Dimensional Order Statistic
This work addresses a specific computational bottleneck in statistics for researchers dealing with score combination, but it appears incremental as it builds on existing order statistic methods.
The paper tackles the problem of combining multiple lists of scores that cannot be directly compared by using joint CDFs of order statistics, and presents a new algorithm with linear runtime in the size of the combined list.
Order statistics provide an intuition for combining multiple lists of scores over a common index set. This intuition is particularly valuable when the lists to be combined cannot be directly compared in a sensible way. We describe here the advantages of a new method for using joint CDFs of such order statistics to combine score lists. We also present, with proof, a new algorithm for computing such joint CDF values, with runtime linear in the size of the combined list.