traceQIsamp: Finding a new evaluation point

In qle: Simulation-Based Quasi-Likelihood Estimation

Description

Minimizing the average total prediction variance of the quasi-score vector

Usage

traceQIsamp(x0, qsd, QD, Sig0 = NULL, nsim = 1000, type = 0L,
  fit = FALSE, pmin = 0.25, ..., optInfo = FALSE, cl = NULL,
  pl = 0L, verbose = FALSE)

Arguments

x0	numeric (named) vector or list as the center point of sampling region, e.g., an estimate of the model parameter
qsd	object of class QLmodel
QD	list of criterion function results, either from calling quasiDeviance or mahalDist
Sig0	sample covariance matrix of parameters given in QD, default NULL, which is then computed
nsim	sample size of random points for computing the integral
type	integer, type=0 (default), use uniform sampling of points, type=1 multivariate normal random points
fit	logical, FALSE (default), whether to re-fit kriging covariance models of the statistics
pmin	minimum required ratio of points falling inside the overall design region (hyperbox of parameters)
...	optional, arguments passed to quasiDeviance, center point 'theta' and 'W' as weighting matrix for kernel estimate of covariance matrix, cross-validation models 'cvm', see quasiDeviance for details
optInfo	logical, FALSE (default), not yet used argument (ignored)
cl	cluster object, NULL (default), of class MPIcluster, SOCKcluster, cluster
pl	print level, use pl>0 to print intermediate results
verbose	if TRUE (default), print intermediate output

Details

Based on the list of sampling candidates QD one point at a time is (sequentially) added to the current sampling design. The integrated (averaged) total prediction variance of the quasi-score vector is then computed based on this new design over a randomly generated set of candidate evaluation points with a sample covariance derived from all sampled candidates. The integral is taken over a discrete set of points either generated uniformly within the design region defined by simple bound constraints if 'Sig0' is NULL or multivariate normally distributed points with covariance matrix 'Sig0' and mean parameter vector 'theta0'.

Value

list of the criterion function value corresponding to the model parameter which minimizes the integrated total prediction variance of the quasi-score vector

Author(s)

M. Baaske

See Also

quasiDeviance #

qle documentation built on May 2, 2019, 5:26 p.m.