traceQIconstr: Find a new local evaluation point

In qle: Simulation-Based Quasi-Likelihood Estimation

Description

The function maximizes the total prediction variance of the quasi-score vector in the vicinity of the center point 'x0' (usually the current parameter estimate). More specifically, a distance weighted version of the trace of the quasi-information matrix is maximized over a local region defined by a level set of hight 'b' w.r.t. the criterion function (either quasi-deviance or Mahalanobis distance) based on the observed statistics. The function can be used to find a suitable next evaluation point. As one of the global selection criteria the function can be called by qle to find a new evaluation point if testing approximate roots is enabled.

Usage

traceQIconstr(x0, qsd, b, method = c("nloptr"), opts = list(), ...,
  optInfo = FALSE, check = TRUE, pl = 0L, verbose = FALSE)

Arguments

x0	(named) numeric vector, the starting and center point of the feasible region
qsd	object of class QLmodel
b	upper bound on either quasi-deviance or mahalanobis distance criterion function values as an upper bound constraint (as a level set function) which defines the feasible region
method	names of possible minimization routines, currently only "nloptr" is supported (see details)
opts	list of control arguments passed to nloptr routines defined in method for global search of maximum and minimum value of quasi-deviance
...	further arguments passed to covarTx
optInfo	logical, FALSE (default), not yet used argument (ignored)
check	logical, TRUE (default), whether to check input arguments
pl	numeric value (>=0), the print level
verbose	if TRUE (default), print intermediate output

Value

list which contains the new evaluation point as the maximizer and the corresponding value of the criterion function

Author(s)

M. Baaske

See Also

searchMinimizer

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