Procedures#

CompML includes frequentist and Bayesian procedures for ML algorithm comparison.

Frequentist tests#

  • Normality: DAgostinoPearsonTest, KolmogorovSmirnovTest, ShapiroWilkTest

  • Homoscedasticity: BartlettTest, LeveneTest

  • Parametric: CorrelatedTTest, PairedTTest, RepeatedMeasuresANOVATest

  • Non-parametric: MannWhitneyUTest, SignTest, SignedRankTest, FriedmanTest, FriedmanAlignedRanksTest, KruskalWallisTest, QuadeTest

Frequentist post-hoc#

  • BonferroniPostHoc

  • FinnerPostHoc

  • HochbergPostHoc

  • HollandPostHoc

  • HolmPostHoc

  • HommelPostHoc

  • LiPostHoc

  • NemenyiPostHoc

  • RomPostHoc

  • ShafferPostHoc

Bayesian tests and post-hoc#

  • BayesianCorrelatedTTest

  • BayesianSignTest

  • BayesianSignedRankTest

  • BayesianHierarchicalTest

  • BayesianFriedmanTest

  • BayesianPostHoc

For signatures and detailed behavior, use the API reference.