Metropolis Package Paclet SymbolMore About »

BlockAdaptiveMetropolis

BlockAdaptiveMetropolis[logpdf, x0, size, stepsize, adpfraction, adprate, adpset]
generates size updates for a list of conditional indepdent random variables starting from x0. Each update is given by the last update of stepsize BlockStudentTMetropolis updatings. The target distribution p[x] is provided by logpdf[x, indices] = Log[p[x, indices]] for any subset x[[indices]] of x. The symmetric proprosal distribution is the student T distribution. The variance of the proposal distribution is adapted in the initial Floor[size adpfraction] iterations with rate adprate, and the adaptive parameter set adpset.
  • To use BlockAdaptiveMetropolis, you first need to load the Metropolis package using Needs["Metropolis`"].
  • Each update is given by {x, logpdf[x], 1/0 (accept/not), Min[1,Metropolis-Hastings ratio]}.