rjpdmp-package: rjpdmp-package

In rjpdmp: Reversible Jump PDMP Samplers

rjpdmp-package

Description

Implements various reversible jump piecewise-deterministic Markov Process methods including the ZigZag and Bouncy Particle Sampler with Normal or Spherical velocity distributions (Chevallier, Fearnhead, Sutton 2020, https://arxiv.org/abs/2010.11771).

The package can be used to Generates PDMP trajectories for reversible jump

• zigzag_logit: ZigZag on logistic likelihood problem

• zigzag_logit_ss: ZigZag with subsampling on Logistic likelihood problem

• bps_s_logit: BPS with velocities distributed uniformly on the sphere for a Logistic likelihood problem

• bps_n_logit: BPS with velocities distributed Normally for a Logistic likelihood problem

• zigzag_rr: ZigZag on a robust regression likelihood problem

• bps_s_rr: BPS with velocities distributed uniformly on the sphere for a robust regression likelihood problem

• bps_n_rr: BPS with velocities distributed Normally for a robust regression likelihood problem

Additional functions

Additional functions for plotting, generating samples, calculating posterior means or probabilities of inclusion

• plot_pdmp: Plot marginal densities and joint pairs plots for trajectories and samples of PDMP samplers and optionally MCMC samples for comparison.

• plot_pdmp_multiple: Plots to compare PDMP samplers and optionally MCMC samples.

• gen_sample: Get samples from PDMP trajectories taking a fixed time discretisation.

• model_probabilities: Calculate either marginal probabilities of inclusions or posterior probabilities of specific models.

• models_visited: Count the number of times a model is visited

• marginal_mean: Calculate the marginal mean using PDMP trajectories

• cond_mean: Calculate the mean conditioned on being in a specific model

Extensions to the package are planned.

rjpdmp documentation built on March 18, 2022, 7:52 p.m.