Computes predictions and the analytical gradient for every observation
using the nlm C machinery (a single multithreaded nlmixr2est::nlmerSolveGrad()
call against the model kept resident by nlmixr2est::.nlmSetupEnv()).
Parameters arrive in positional order matching
.nlmerGlobal$nlmerEnv$paramNames (THETA[i] order).
Usage
.nlmerNlmerFun(TIME, ...)
Arguments
- TIME
Observation times vector (unused; kept for formula
compatibility – the solver uses the resident per-subject event data)
- ...
Individual-level parameter vectors (one per param in
saemParamsToEstimate)
Value
Numeric vector of predictions with "gradient" attribute
Details
lme4 passes one value per observation for each nonlinear parameter; within a
subject these are constant (phi = beta + b), so the per-subject parameter
vector is read from each subject's first observation row.
