Wilcoxon

From Eigenvector Research Documentation Wiki
Jump to navigation Jump to search

Purpose

Pairwise Wilcoxon signed rank test for evaluating residuals from two models.

Synopsis

prob = wilcoxon(err_1,err_2)

Description

Pairwise comparison between two sets of model residuals using the sign and magnitude of residuals from two models. Output is the probability that model 2 (the model producing the second set of residuals) is better than model 1 (the model that produces the first set of residuals).

Adapted from: Edward V. Thomas, "Non-parametric statistical methods for multivariate calibration model selection and comparison", J. Chemometrics 2003; 17: 653–659. Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/cem.833

Inputs

  • err_1 = Prediction errors from model #1
  • err_2 = Prediction errors from model #2


Outputs

  • prob = Prob{# of times model#2 wins <=k} Probability that model#2 is better than model#1.

See Also

crossval, randomttest, signtest