adapump: Illustration of the danger of adaptive MCMC for the pump failure data
Description
This function constructs an update of the location and scale matrix of a normal proposal
in a Metropolis-Hastings algorithm, based on earlier simulations
in order to show the danger of online adaptivity.
Usage
adapump(T = 10^2, MM = 10^2)
Arguments
T
Number of steps between updates
MM
Number of updates, leading to a chain of length MM*T
Value
The function simply plots the sequence of $beta$'s along iterations, which should
collapse, as well as the range of the variability of the proposed values.
References
Chapter 8 of EnteR Monte Carlo Statistical Methods