EI: Expected improvement

In depCensoring: Statistical Methods for Survival Data with Dependent Censoring

Expected improvement

Description

Used in the M-step (M_step.R). Note: predict(fit.krige, ...) has weird beheviour when making predictions for a single value in terms of standard error. We work around this issue in this implementation.

Usage

EI(theta, test.fun, fit.krige, theta.hash, dir)

Arguments

theta	Vector of coefficients.
test.fun	Test function (cf. EstimationAlgorithmBei.R).
fit.krige	Fitted Kriging model.
theta.hash	Tentative optimal value for theta, i.e., the largest or smallest feasible value for theta (if dir = 1 or dir = -1, respectively). A 'feasible value' is one that satisfies all moment restrictions.
dir	Search direction. dir = 1 corresponds to looking for an upper bound. dir = -1 corresponds to looking for a lower bound.

Value

The expected improvement.

depCensoring documentation built on April 4, 2025, 1:52 a.m.