mhmix: Implement two Metropolis-Hastings algorithms on a mixture posterior
Description
This function runs a Metropolis-Hastings algorithm on a posterior distribution associated with a
mixture model and 500 datapoints. Depending on the value of the boolean indicator lange, the function
uses a regular Gaussian random-walk proposal or a Langevin alternative.
Usage
mhmix(Niter = 10^4, lange = FALSE, scale = 1)
Arguments
Niter
Number of MCMC iterations
lange
Boolean variable indicating the use of the Langevin alternative
scale
Scale factor of the Gaussian perturbation
Value
The function returns a plot of the log-posterior surface, along with the MCMC sample represented
both by points and lines linking one value to the next.
References
Chapter 6 of EnteR Monte Carlo Statistical Methods