KLopt.lnorm: Calculation of KL-optimal discriminating design for lognormal... In rodd: Optimal Discriminating Designs

Description

Calculates an approximation xi^{**} of the KL-optimal design (in case of lognormal errors) xi^* for discrimination between a given list of error densities {f_i(x,theta_i), i = 1,…,nu}. This procedure is based on the work [8]. This function mimics tpopt almost entirely. It is planed to combine tpopt and KLopt.lnorm in the future. See tpopt for the detailed description of the arguments marked with “-//-”.

Usage

 1 2 3 4 5 6 7 8 9 10 KLopt.lnorm( x, w = rep(1, length(x)) / length(x), eta, sq.var, theta.fix, theta.var = NULL, p, x.lb = min(x), x.rb = max(x), opt = list())

Arguments

 x -//- w -//- eta a list of means for the error densities {f_i(x,theta_i), i = 1,…,nu} between which proposed optimization should be performed. Every function from this list should be defined in the form of eta_i(x,theta_i), where x is one dimensional variable from X and θ_i is a vector of corresponding model parameters. We will refer to length of this list as nu. sq.var a list of variances for the error densities {f_i(x,theta_i), i = 1,…,nu} between which proposed optimization should be performed. Every function from this list should be defined in the form of v^2_i(x,theta_i). This list also has the length equal to nu. theta.fix -//- theta.var -//- p -//- x.lb -//- x.rb -//- opt -//-

Value

Object of class “KLopt.lnorm” which contains the following fields:

x, w, efficiency, functional

-//-

eta

a list of means from the input.

sq.var

a list of variances from the input.

theta.fix, theta.var, p, x.lb, x.rb, max.iter, done.iter, des.eff, time

-//-