DrawInit function

Random draws for initialization

Utility function for the package EntropyMCMC, for generating random starting positions for the parallel Markov chains, used by, e.g., MCMCcopies or EntropyParallel.

DrawInit(nmc, d, initpdf="rnorm", ...)

Arguments

• nmc: Number of parallel chains = initial points.
• d: Space dimension.
• initpdf: Random generator. Generators currently implemented are: "rnorm" as the Normal distribution and "runif" as the uniform distribution.
• ...: Parameters passed to initpdf

Returns

DrawInit returns a matrix of dimension (nmc,d) where each row is a $d$-dimensional point.

Author(s)

Didier Chauveau.

See Also

MCMCcopies and MCMCcopies.mc for iid MCMC simulations, EntropyParallel and EntropyParallel.cl

for simultaneous simulation and entropy estimation.

Note

It is better for mixing properties to use diffuse initial distributions, such as the one proposed here. However Dirac initial points can also be used, precisely to evaluate the efficiency of a MCMC to escape from a wrong initial position (e.g., in the tails of the target density).

Examples

Ptheta0 <- DrawInit(10, 5, initpdf="rnorm", mean=0, sd=5)