Python Model Selection Toolkit (spelled PyMiSTake)

A small Python toolkit for the estimation of global likelihoods (Bayesian evidence).

Currently implemented routines

• Bridge sampling with a proxy distribution
  • Estimates the absolute global likelihood given
    • a set of posterior sample locations
    • a set of posterior samples corresponding to the locations
    • posterior distribution function
    • proxy distribution (optional)

Routines to be implemented

• Nested sampling
• Truncated posterior mixture estimate

Trivial routines that may be implemeted

• Basic MC integration
• Importance sampling MC integration

Nontrivial routines that may (or may not) be implemented

• Bayesian quadrature

• Hannu Parviainen hpparvi@gmail.com