nlmObjectiveSetup: Set up an nlm-family objective for repeated hook-firing...

In nlmixr2est: Nonlinear Mixed Effects Models in Population PK/PD, Estimation Routines

Set up an nlm-family objective for repeated hook-firing evaluation

Description

Preprocesses the data and LOADS the nlm population (predOnly) problem into the C++ engine, so that repeated nlmSolveR(theta) calls evaluate the population objective – firing any registered likelihood-contribution hook (e.g. a plugin's per-observation cotangent capture) – WITHOUT re-running the optimizer. One compiled setup is reused across evaluations. Intended for extension packages (e.g. nlmixr2nn) that optimize an out-of-band parameter block (network weights injected via a par-loader) and need the exact error-model cotangent from the nlm C++ solve at each iterate. Free the loaded problem with .nlmFreeEnv() when done.

Usage

nlmObjectiveSetup(
  ui,
  data,
  control = NULL,
  gradient = FALSE,
  scale = c("control", "natural")
)

Arguments

ui	rxode2/nlmixr2 model. Uses ui$control when present.
data	event data.
control	optional nlm-family control; defaults to nlmControl() (or ui$control if that is an nlm-family control).

Value

(invisibly) the scaled starting parameter vector to hand to nlmSolveR(); the C++ problem is left loaded.

Author(s)

Matthew L. Fidler

nlmixr2est documentation built on Sept. 20, 2026, 9:08 a.m.