EasyDeepLearn

WAIC and LOO-CV — Bayesian model comparison.

hard

Answer

  • Both estimate out-of-sample predictive performance.
  • WAIC = -2    (lppd    pWAIC)2\; \cdot \;(\mathrm{lppd}\; - \;p_{\mathrm{WAIC}}) where lppd = expected log posterior predictive on training data, pWAIC  =  effectivep_{\mathrm{WAIC}}\; = \;\mathrm{effective} number of parameters via posterior variance.
  • LOO-CV: exact leave-one-out log predictive; approximated efficiently via PSIS-LOO (Vehtari et al.) using importance sampling — modern default in Stan / PyMC / Arviz.
  • Both replace AIC / BIC / Bayes factor for model comparison in most cases.
Check yourself — multiple choice
  • Only AIC
  • Both estimate out-of-sample predictive fit; WAIC uses lppd  +  pWAIC\mathrm{lppd}\; + \;p_{\mathrm{WAIC}}; PSIS-LOO approximates LOO-CV via importance sampling — modern Stan/PyMC default
  • Same as BIC
  • Not real

WAIC / PSIS-LOO: modern predictive-fit Bayesian model comparison.

#bayesian

Practise Statistics Fundamentals

215 interview questions in this topic.

Related questions