metropolis function

Metropolis sampling of a continuous distribution

Implements Metropolis sampling for an arbitrary continuous probability distribution

metropolis(logpost, current, C, iter, ...)

Arguments

• logpost: function definition of the log probability function
• current: starting value of algorithm
• C: half-width of proposal interval
• iter: number of iterations
• ...: other inputs needed in logpost function

Returns

• S: vector of simulated values

• accept_rate: acceptance rate of algorithm

Author(s)

Jim Albert

Examples

lpost <- function(theta, s){
  dnorm(s$ybar, theta, s$se, log = TRUE) +
    dcauchy(theta, s$loc, s$scale, log = TRUE)
}
s <- list(ybar = 20,
          se = 0.4,
          loc = 10,
          scale = 2)
post <- metropolis(lpost, 10, 20, 100, s)