metropolis: Metropolis algorithm

In rpruim/CalvinBayes: Support for Computational Bayesian Analysis Courses

Description

Simple Metropolis algorithm implementations for simple models.

Usage

metro_bern(
  x,
  n,
  step_size = 0.01,
  start = 0.5,
  num_steps = 10000,
  prior = dunif,
  ...
)

metro_norm(
  y,
  num_steps = 1e+05,
  step_size = 1,
  start = list(mu = 0, log_sigma = 0)
)

Arguments

x	number of success in data
n	number of failures in data
step_size	sd of (normal) jump distribution(s)
start	starting value(s) for MCMC
num_steps	how long to run MCMC
prior	a function describing the prior
...	additional arguments for the prior
y	vecctor of numeric response values

Examples

Metro <-
  metro_bern(10, 30, step_size = 0.1, prior = dbeta, shape1 = 4, shape2 = 4)
# posterior density
Metro %>%
  gf_dens(~ theta) %>%
  gf_dist("beta", shape1 = 14, shape2 = 24, color = "red")
# trace plot
Metro %>%
  gf_line(theta ~ step)

metro_norm(rnorm(25, 10, 1), start = list(mu = 5, log_sigma = log(5))) %>%
  gf_density( ~ mu)

rpruim/CalvinBayes documentation built on April 12, 2021, 1:49 p.m.