bayesdistreg (version 0.1.0)

IndepMH: Independence Metropolis-Hastings Algorithm

Description

IndepMH computes random draws of parameters using a specified proposal distribution.

Usage

IndepMH(data, propob = NULL, posterior = NULL, iter = 1500,
  burn = 500, vscale = 1.5, start = NULL, prior = "Uniform",
  mu = 0, sig = 10)

Arguments

data

data required for the posterior distribution

propob

a list of mean and variance-covariance of the normal proposal distribution (default:NULL)

posterior

the posterior distribution. It is set to null in order to use the logit posterior. The user can specify log posterior as a function of parameters and data (pars,data)

iter

number of random draws desired (default: 1500)

burn

burn-in period for the MH algorithm (default: 500)

vscale

a positive value to scale up or down the variance-covariance matrix in the proposal distribution

start

starting values of parameters for the MH algorithm. It is automatically generated but the user can also specify.

prior

the prior distribution (default: "Normal", alternative: "Uniform")

mu

the mean of the normal prior distribution (default:0)

sig

the variance of the normal prior distribution (default:10)

Value

val a list of matrix of draws pardraws and the acceptance rate

Examples

# NOT RUN {
y = indicat(faithful$waiting,70)
x = scale(cbind(faithful$eruptions,faithful$eruptions^2))
data = data.frame(y,x); propob<- lapl_aprx(y,x)
IndepMH_n<- IndepMH(data=data,propob,iter = 102, burn = 2) # prior="Normal"
IndepMH_u<- IndepMH(data=data,propob,prior="Uniform",iter = 102, burn = 2) # prior="Uniform"
par(mfrow=c(3,1));invisible(apply(IndepMH_n$Matpram,2,function(x)plot(density(x))))
invisible(apply(IndepMH_u$Matpram,2,function(x)plot(density(x))));par(mfrow=c(1,1))

# }