trim: Trim/Thin MCMC iterations

In deepgp: Bayesian Deep Gaussian Processes using MCMC

Trim/Thin MCMC iterations

Description

Acts on a gp, gpvec, dgp2, dgp2vec, dgp3vec, or dgp3 object. Removes the specified number of MCMC iterations (starting at the first iteration). After these samples are removed, the remaining samples are optionally thinned.

Usage

trim(object, burn, thin)

## S3 method for class 'gp'
trim(object, burn, thin = 1)

## S3 method for class 'gpvec'
trim(object, burn, thin = 1)

## S3 method for class 'dgp2'
trim(object, burn, thin = 1)

## S3 method for class 'dgp2vec'
trim(object, burn, thin = 1)

## S3 method for class 'dgp3'
trim(object, burn, thin = 1)

## S3 method for class 'dgp3vec'
trim(object, burn, thin = 1)

Arguments

object	object from fit_one_layer, fit_two_layer, or fit_three_layer
burn	integer specifying number of iterations to cut off as burn-in
thin	integer specifying amount of thinning (thin = 1 keeps all iterations, thin = 2 keeps every other iteration, thin = 10 keeps every tenth iteration, etc.)

Details

The resulting object will have nmcmc equal to the previous nmcmc minus burn divided by thin. It is recommended to start an MCMC fit then investigate trace plots to assess burn-in. Once burn-in has been achieved, use this function to remove the starting iterations. Thinning reduces the size of the resulting object while accounting for the high correlation between consecutive iterations.

Value

object of the same class with the selected iterations removed

Examples

# See ?fit_one_layer, ?fit_two_layer, or ?fit_three_layer
# for examples

deepgp documentation built on Sept. 11, 2024, 8:30 p.m.