Title: A sensitivity analysis for choosing reliability functions to select cladograms

(2006) A sensitivity analysis for choosing reliability functions to select cladograms. Zoologica scripta, 00, 000-000. Different methods to select among reliability functions used in cladistics (linear parsimony and implied weights) are proposed and explored under a sensitivity analysis framework. The proposed methods are character, taxon and mixed jackknife, and fast searches. Although results seems to be matrix-procedure dependent some general behaviors could be extract from the analysis. We found that the absolute number of recover nodes is the best measure for selection, as they are independent of resolution of compared cladograms. Scaled topological measures strongly depend on resolution of compared cladograms. Fit/distortion measures are biased towards the strongest functions of implied weights. Also a small number of replications (as 20) produces nearly the same results than a huge number of replicates. Implied weights outperform linear parsimony in terms of stability of results. Given those results, perspectives and analysis guidelines are proposed.