Function, option and example index

This chapter lists every ferx-r function, setting, model file block, data column, bundled example, warning category and fit object slot in the pinned build (6e2f701), with the chapter that covers it. It is generated from the book’s feature inventory, the package’s help pages and the ferx-core documentation of the pinned engine.

Functions

Help pages: ?name. Function reference on the web: ferx-r reference.

Function	Description	Arguments	Chapter
`check_diagnostics()`	Structured diagnostic flags from a fit result	`fit`	Initial estimates and a first fit
`check_strictness()`	Eligibility gates for a candidate fit	`fit`, `require_converged = TRUE`, `require_covariance = FALSE`, `max_condition_number = 1000`, `max_correlation = 0.95`, `reject_on_boundary = TRUE`, `reject_init_stall = TRUE`	Model development and selection
`ferx_allometry()`	Allometric scaling	`model = NULL`, `data = NULL`, `config = NULL`, `covariate = NULL`, `reference = NULL`, `parameters = NULL`, `exponents = NULL`, `estimate = FALSE`, `lower = NULL`, `upper = NULL`, `fit = TRUE`, `threads = NULL`, `retries = NULL`, `directory = NULL`	Covariate modeling
`ferx_amd()`	Automatic model development (AMD)	`model = NULL`, `data = NULL`, `search_space = NULL`, `config = NULL`, `strategy = NULL`, `retries_on = NULL`, `skip = NULL`, `rank = NULL`, `cutoff = NULL`, `threads = NULL`, `retries = NULL`, `directory = NULL`, `resume = FALSE`, `progress`	Model development and selection
`ferx_amd_plan()`	The AMD pipeline as planned, without fitting anything	`model = NULL`, `data = NULL`, `search_space = NULL`, `config = NULL`, `strategy = NULL`, `retries_on = NULL`, `skip = NULL`	Model development and selection
`ferx_apply_selection()`	Preview data-selection filtering	`data`, `ignore = NULL`, `accept = NULL`, `ignore_ids = NULL`, `excluded = FALSE`	Preparing and checking the analysis dataset
`ferx_bic()`	BIC variants for model selection	`fit`, `type`	Model development and selection
`ferx_bootstrap()`	Non-parametric case bootstrap	`model`, `data = NULL`, `samples = 200`, `seed = 1`, `threads = NULL`, `stratify_on = NULL`, `sample_size = NULL`, `update_inits = TRUE`, `run_base_model = TRUE`, `keep_covariance = FALSE`, `dofv = FALSE`, `skip_minimization_terminated = TRUE`, `skip_estimate_near_boundary = TRUE`, `skip_covariance_step_terminated = FALSE`, `skip_with_covstep_warnings = FALSE`, `ci = 95`, `directory = NULL`, `resume = FALSE`, `retry_failed = FALSE`, `progress`, `verbose = FALSE`	Parameter uncertainty
`ferx_bootstrap_summarize()`	Re-summarise a finished bootstrap run	`directory`, `skip_minimization_terminated = TRUE`, `skip_estimate_near_boundary = TRUE`, `skip_covariance_step_terminated = FALSE`, `skip_with_covstep_warnings = FALSE`, `ci = 95`	Parameter uncertainty
`ferx_calc_npde()`	Simulation-based NPDE / NPD diagnostics from a fit	`fit`, `nsim = 1000L`, `seed = NULL`, `model = NULL`, `data = NULL`	Diagnosing the model
`ferx_check_init()`	Quick convergence check with an optimizer trace	`model`, `data = NULL`, `method = "focei"`, `maxiter = NULL`	Initial estimates and a first fit
`ferx_coef()`	Pull a parameter estimate or its standard error by name	`fit`, `param = NULL`	Initial estimates and a first fit
`ferx_collect()`	Retrieve the result of a background fit	`handle`, `verbose = TRUE`	Estimation methods and controlling the fit
`ferx_conddist()`	Per-subject SAEM conditional distribution of random effects	`fit`	Estimation methods and controlling the fit
`ferx_cov_screen()`	Quick covariate screen against individual parameters and ETAs	`fit`, `threshold = 0.2`	Covariate modeling
`ferx_covariance()`	Run the covariance step against an existing fit	`fit`, `covariance_method = "r"`, `mu_referencing = TRUE`, `verbose = FALSE`	Parameter uncertainty
`ferx_covsearch()`	Stepwise covariate modelling (SCM)	`model = NULL`, `data = NULL`, `search_space = NULL`, `config = NULL`, `algorithm = NULL`, `p_forward = NULL`, `p_backward = NULL`, `max_steps = NULL`, `adaptive_scope_reduction = NULL`, `rank = NULL`, `cutoff = NULL`, `threads = NULL`, `retries = NULL`, `directory = NULL`, `resume = FALSE`, `progress`	Model development and selection
`ferx_example()`	Get paths to example model and data files	`name = NULL`	A complete analysis in one chapter
`ferx_fit()`	Fit a nonlinear mixed effects model	`model`, `data = NULL`, `method = NULL`, `covariance = NULL`, `verbose = NULL`, `bloq_method = NULL`, `threads = NULL`, `mu_referencing = NULL`, `sir = NULL`, `gradient = NULL`, `optimizer_trace = FALSE`, `scale_params = FALSE`, `inits_from_nca = FALSE`, `fd_hessian_step = NULL`, `settings = NULL`, `output = NULL`, `include_data = FALSE`, `ignore = NULL`, `accept = NULL`, `ignore_ids = NULL`	Initial estimates and a first fit
`ferx_fit_async()`	Fit a model asynchronously	`model`, `data = NULL`, `tail_n = 6L`, `poll_interval = 0.5`	Estimation methods and controlling the fit
`ferx_gam_screen()`	GAM-based covariate pre-screening	`fit`, `etas = NULL`, `covariates = NULL`, `spline_df`, `include_linear = TRUE`, `shrinkage_warn = 0.3`	Covariate modeling
`ferx_get_columns()`	Inspect column headers of a data file	`data`	Preparing and checking the analysis dataset
`ferx_get_warnings()`	Structured warnings from a ferx fit	`fit`, `as_df = FALSE`	Initial estimates and a first fit
`ferx_globalsearch()`	Global model search	`model = NULL`, `data = NULL`, `search_space = NULL`, `config = NULL`, `algorithm = NULL`, `iiv_strategy = NULL`, `max_models = NULL`, `ga = NULL`, `penalties = NULL`, `rank = NULL`, `cutoff = NULL`, `threads = NULL`, `retries = NULL`, `directory = NULL`, `resume = FALSE`, `progress`	Model development and selection
`ferx_iivsearch()`	Variability-structure search	`model = NULL`, `data = NULL`, `search_space = NULL`, `config = NULL`, `algorithm = NULL`, `correlation_algorithm = NULL`, `as_fullblock = NULL`, `block_retries = NULL`, `rank = NULL`, `cutoff = NULL`, `threads = NULL`, `retries = NULL`, `directory = NULL`, `resume = FALSE`, `progress`	Model development and selection
`ferx_inits_from_nca()`	Derive NCA-based starting values without fitting	`model`, `data = NULL`, `method`	Initial estimates and a first fit
`ferx_iovsearch()`	Inter-occasion variability search	`model = NULL`, `data = NULL`, `search_space = NULL`, `config = NULL`, `column = NULL`, `distribution = NULL`, `groups = NULL`, `block_retries = NULL`, `rank = NULL`, `cutoff = NULL`, `threads = NULL`, `retries = NULL`, `directory = NULL`, `resume = FALSE`, `progress`	Model development and selection
`ferx_load_fit()`	Load a fit from a .fitrx bundle	`path`	Reproducibility and sharing
`ferx_model()`	Create a ferx_model object	`data = NULL`, `model = NULL`, `template = NULL`, `path = NULL`, `overwrite = FALSE`, `edit = TRUE`, `print = FALSE`	Writing and managing model files
`ferx_model_edit()`	Open a ferx model file in an editor	`path`, `dest = "."`, `overwrite = FALSE`, `save_as = NULL`, `.editor`	Writing and managing model files
`ferx_model_get_section()`	Get a section from a ferx model	`x`, `section`, `strip = FALSE`	Writing and managing model files
`ferx_model_inspect()`	Inspect the structure of a ferx model file	`path`	Writing and managing model files
`ferx_model_set_section()`	Set a section in a ferx model	`x`, `section`, `lines`	Writing and managing model files
`ferx_model_show()`	Display a ferx model file in the console	`path`	Writing and managing model files
`ferx_model_to_frem()`	Prepare a FREM (Full Random Effects Model) dataset and model	`model`, `data = NULL`, `covariates = NULL`, `output_dir = NULL`, `output_model = NULL`, `output_data = NULL`, `fit = NULL`	Covariate modeling
`ferx_model_validate()`	Validate a ferx model file for syntax and structural errors	`path`, `data = NULL`	Writing and managing model files
`ferx_modelsearch()`	Structural PK model search	`model = NULL`, `data = NULL`, `search_space = NULL`, `config = NULL`, `algorithm = NULL`, `iiv_strategy = NULL`, `rank = NULL`, `cutoff = NULL`, `threads = NULL`, `retries = NULL`, `directory = NULL`, `resume = FALSE`, `progress`	Model development and selection
`ferx_predict()`	Population predictions from a NLME model	`model`, `data = NULL`, `fit = NULL`	Simulating scenarios
`ferx_predict_survival()`	Survival-function predictions for time-to-event endpoints	`model`, `data = NULL`, `times`, `fit = NULL`	Time-to-event models
`ferx_runlog()`	Run log for a ferx fit	`fit`, `gradient_tol = 0.1`, `show_iterations = TRUE`, `verbose = TRUE`	Estimation methods and controlling the fit
`ferx_runlog_iters()`	Print the full optimizer iteration history for a ferx fit	`fit`, `verbose = TRUE`	Estimation methods and controlling the fit
`ferx_ruvsearch()`	Residual-error model search	`model = NULL`, `data = NULL`, `config = NULL`, `groups = NULL`, `p_value = NULL`, `skip = NULL`, `max_iter = NULL`, `cwres_prescreen = NULL`, `threads = NULL`, `retries = NULL`, `directory = NULL`, `resume = FALSE`, `progress`	Model development and selection
`ferx_save_fit()`	Save a ferx fit to a portable .fitrx bundle	`fit`, `output`, `include_data = FALSE`	Reproducibility and sharing
`ferx_se()`	Pull a parameter estimate or its standard error by name	`fit`, `param = NULL`	Initial estimates and a first fit
`ferx_search_config()`	Load and validate a .ferxsearch search configuration	`path`	Model development and selection
`ferx_search_coverage()`	Coverage of a search space against what the engine can express	`space`	Model development and selection
`ferx_search_results()`	Read a search run’s table	`directory`, `partial = NULL`, `type`	Model development and selection
`ferx_search_space()`	Parse and expand an MFL search space	`mfl`, `model = NULL`, `data = NULL`	Model development and selection
`ferx_simulate()`	Simulate from a NLME model	`model`, `data = NULL`, `n_sim = 1L`, `seed = 42L`, `fit = NULL`, `match = FALSE`, `horizon = NULL`	Simulation-based evaluation: VPC
`ferx_simulate_adaptive()`	Simulate state-reactive (adaptive / feedback) dosing	`model`, `data = NULL`, `n_sim = 1L`, `seed = 42L`, `verify = TRUE`, `max_decisions = 0L`	Adaptive dosing and TDM strategies
`ferx_simulate_with_uncertainty()`	Simulate with parameter-uncertainty propagation	`model`, `data`, `fit`, `n_uncertainty_draws = 100L`, `n_sim_per_draw = 1L`, `method`, `seed = 42L`	Simulating scenarios
`ferx_sir()`	Run SIR against an existing fit	`fit`, `sir_samples = 1000L`, `sir_resamples = 250L`, `sir_seed = NULL`, `sir_keep_samples = FALSE`, `verbose = FALSE`	Parameter uncertainty
`ferx_stop()`	Stop a background fit	`handle`	Estimation methods and controlling the fit
`ferx_trace()`	Read the optimizer trace from a ferx fit	`fit`	Estimation methods and controlling the fit
`ferx_xpose()`	Build an object from a ferx fit	`fit`, `backend`, `continuous = NULL`, `categorical = NULL`, `runno = 1L`, `iterations = TRUE`	Diagnosing the model

S3 methods

Method	Chapter
`$.ferx_job`	Estimation methods and controlling the fit
`plot.ferx_bootstrap`	Parameter uncertainty
`plot.ferx_fit`	Estimation methods and controlling the fit
`plot.ferx_job`	Estimation methods and controlling the fit
`print.ferx_allometry`	Covariate modeling
`print.ferx_amd`	Model development and selection
`print.ferx_bootstrap`	Parameter uncertainty
`print.ferx_conddist`	Estimation methods and controlling the fit
`print.ferx_covsearch`	Model development and selection
`print.ferx_data`	Preparing and checking the analysis dataset
`print.ferx_fit`	Initial estimates and a first fit
`print.ferx_globalsearch`	Model development and selection
`print.ferx_iivsearch`	Model development and selection
`print.ferx_inits`	Initial estimates and a first fit
`print.ferx_iovsearch`	Model development and selection
`print.ferx_job`	Estimation methods and controlling the fit
`print.ferx_model`	Writing and managing model files
`print.ferx_modelsearch`	Model development and selection
`print.ferx_ruvsearch`	Model development and selection
`print.ferx_search_config`	Model development and selection
`print.ferx_search_space`	Model development and selection
`print.ferx_summary`	Initial estimates and a first fit
`summary.ferx_amd`	Model development and selection
`summary.ferx_covsearch`	Model development and selection
`summary.ferx_fit`	Initial estimates and a first fit
`summary.ferx_globalsearch`	Model development and selection
`summary.ferx_iivsearch`	Model development and selection
`summary.ferx_iovsearch`	Model development and selection
`summary.ferx_modelsearch`	Model development and selection
`summary.ferx_ruvsearch`	Model development and selection

Fit settings

Keys for ferx_fit(settings = list(...)) and the [fit_options] block. The chapter column points at the section that uses the key; the ferx-core fit options page gives the full semantics. A blank Values cell means the key takes a free value (a number, a name or a list) rather than a fixed set; a blank Meaning means the pinned engine documents the key only in terms of other software, which this book does not quote.

Setting	Values	Default	Meaning	Chapter
`accept`			Character vector of filter expressions. A record is kept only when all conditions pass; excluded if any fail.	Preparing and checking the analysis dataset
`adapt_interval`		`50`	Iterations between step-size adaptation	Estimation methods and controlling the fit
`agq_eval_only`				Estimation methods and controlling the fit
`analytic_cov_hessian`	`true`, `false`	`true`	Use the exact analytic R-matrix for the covariance step when the model is in scope, instead of the finite-difference-of-OFV Hessian.	Parameter uncertainty
`bayes_chains`			Number of independent chains (default 4); used for split-R-hat.	Estimation methods and controlling the fit
`bayes_iters`			Retained sampling sweeps per chain, before thinning (default 1000).	Estimation methods and controlling the fit
`bayes_seed`			Base RNG seed for the Bayes sampler. Independent of seed / saem_seed.	Estimation methods and controlling the fit
`bayes_thin`			Keep every bayes_thin-th sampling draw (default 1).	Estimation methods and controlling the fit
`bayes_warmup`			Warmup (burn-in + adaptation) sweeps per chain, discarded from the posterior (default 1000).	Estimation methods and controlling the fit
`checkpoint`	`true`, `false`	`true`	Periodically save a resume point so a run that is interrupted (e.g. `Ctrl-C`, a crashed job) can be restarted from where it stopped instead of from scratch.	Estimation methods and controlling the fit
`checkpoint_interval_secs`	integer	`300`	Minimum wall-clock seconds between checkpoint writes. Larger values reduce I/O overhead but lose more progress on an interrupt; smaller values save more often.	Estimation methods and controlling the fit
`conddist`		`false`	Run a post-fit conditional-distribution pass estimating each subject’s `p(η_i \\| y_i)` by MCMC.	Estimation methods and controlling the fit
`conddist_burnin`		`20`	Burn-in draws discarded before accumulation, to forget the EBE-mode warm start.	Estimation methods and controlling the fit
`conddist_keep_samples`		`false`	Retain the raw per-subject draws (written to `{model}-conddist-samples.csv`), not just the mean/SD.	Estimation methods and controlling the fit
`conddist_nsamp`		`200`	Retained MCMC draws per subject in the conditional-distribution pass.	Estimation methods and controlling the fit
`cov_inner_tol`	float	`inner_tol` (LTBS: `min(inner_tol, 1e-8)`)	Inner EBE-reconvergence tolerance used only by the covariance step, decoupled from `inner_tol`.	Parameter uncertainty
`covariance_fallback`	`none`, `sir`	`none`	What to do when the covariance Hessian is non-PD, whether `R` is analytic or finite-difference.	Parameter uncertainty
`covariance_method`	`r`, `s`, `rsr`	`r`		Parameter uncertainty
`ebe_warm_start`		`false`	When an inner per-subject EBE solve fails its BFGS step and falls back to Nelder–Mead, warm-start the simplex from the BFGS partial η̂ rather than cold-starting from the prior mode η=0.	Estimation methods and controlling the fit
`fd_hessian_step`	float	`1e-2`	Initial relative step size for the finite-difference Hessian used in the covariance step.	Parameter uncertainty
`frem_predictions`				Covariate modeling
`frem_rao_blackwell`		`true`	FREM only: Rao-Blackwellise the covariate ETAs (integrate them analytically, importance-sample only the PK ETAs).	Covariate modeling
`frem_sigma`				Covariate modeling
`global_maxeval`	integer	`0` (auto: `30 * (n_params + 1)`)	Maximum evaluations of the FOCE objective during the global pre-search.	Estimation methods and controlling the fit
`global_search`	`true`, `false`	`false`	Run NLopt CRS2-LM (Controlled Random Search with Local Mutation) as a gradient-free global pre-search before the local optimizer.	Estimation methods and controlling the fit
`gn_lambda`			Levenberg-Marquardt damping factor (default 0.01). Larger values make steps more conservative.	Estimation methods and controlling the fit
`ignore`			Character vector of filter expressions that exclude a record when the expression evaluates to true.	Preparing and checking the analysis dataset
`ignore_subjects`				Preparing and checking the analysis dataset
`imp_auto`		`true`	Adaptive sample count. When `true`, `imp_samples` is the starting count and is ramped up (×2/iteration, cap 10000) while the objective’s Monte-Carlo SE exceeds 1.0.	Estimation methods and controlling the fit
`imp_averaging`		`50`	Terminal iterations averaged into the reported estimate (estimator only).	Estimation methods and controlling the fit
`imp_defensive_alpha`		`0.0`	Defensive-mixture weight (issue #528), opt-in. Each subject draws this fraction of its samples from the prior `N(0, Ω)` rather than the mode-centred proposal, and every sample is scored under the mixture density.	Estimation methods and controlling the fit
`imp_eval_only`		`false`	`true` ⇒ evaluate `−2 log L` at fixed parameters; must be the terminal chain stage.	Estimation methods and controlling the fit
`imp_iterations`		`200`	MCEM iterations (estimator only).	Estimation methods and controlling the fit
`imp_low_ess_threshold`		`0.1`	Subjects with normalized ESS below this fraction get flagged in the result.	Estimation methods and controlling the fit
`imp_proposal_df`		`5.0`	Student-t proposal degrees of freedom (≥ 1), or `normal`/`mvn` for a multivariate-normal proposal.	Estimation methods and controlling the fit
`imp_samples`		`1000`	Importance samples K per subject. 2000–5000 recommended for publication-quality MC SE.	Estimation methods and controlling the fit
`imp_seed`		`12345`	RNG seed. Same seed → identical result.	Estimation methods and controlling the fit
`impmap_auto`		`true`	Adaptive sample count. When `true`, `impmap_samples` is the starting count and is ramped up (×2/iteration, cap 10000) while the objective’s Monte-Carlo SE exceeds 1.0.	Estimation methods and controlling the fit
`impmap_averaging`		`50`	Final iterations whose parameters are averaged into the reported estimate (Monte-Carlo variance reduction).	Estimation methods and controlling the fit
`impmap_iterations`		`200`	Number of MCEM iterations (parameter updates).	Estimation methods and controlling the fit
`impmap_low_ess_threshold`		`0.1`	Subjects with normalized ESS below this fraction are flagged as poorly sampled.	Estimation methods and controlling the fit
`impmap_mceta`		`0`	Number of additional random starting points for per-subject MAP optimization.	Estimation methods and controlling the fit
`impmap_proposal_df`		`4`	Proposal degrees of freedom. A finite value ≥ 1 gives a heavier-tailed Student-t (default `4`); `normal` (or `mvn`) gives a multivariate-normal proposal.	Estimation methods and controlling the fit
`impmap_samples`		`300`	Importance samples K per subject per iteration. Larger K reduces Monte-Carlo noise at linear cost.	Estimation methods and controlling the fit
`impmap_seed`		`12345`	RNG seed. Same seed → identical estimates.	Estimation methods and controlling the fit
`impmap_sobol`		`false`	Use Sobol quasi-random sequences (with Cranley-Patterson randomization) for IS draws instead of pseudo-random.	Estimation methods and controlling the fit
`impmap_trace`		`false`	Logical; when TRUE, collect per-iteration parameter values into fit$impmap_trace.	Estimation methods and controlling the fit
`inits_from_nca`	`true`, `false`, `nca`, `nca_sweep`, `nca_ebe`	`false`	Derive NCA-based starting values from the data before the optimizer loop.	Initial estimates and a first fit
`inner_maxiter`	integer	`200`	Max iterations for the inner (per-subject EBE) optimizer	Estimation methods and controlling the fit
`inner_optimizer`	`auto`, `bfgs`, `lbfgs`, `nelder_mead`	`auto`	Inner-loop (per-subject EBE) optimisation algorithm. `auto` uses dense BFGS, switching to limited-memory L-BFGS above 32 random effects.	Estimation methods and controlling the fit
`inner_restarts`	integer	`1`	Guarded multi-start count for the inner (per-subject EBE) optimizer, to escape a multimodal individual objective.	Estimation methods and controlling the fit
`inner_tol`	float	`1e-5`	Gradient-norm convergence tolerance for the inner (per-subject EBE) optimizer.	Estimation methods and controlling the fit
`iov_column`	string	—	Name of the occasion column in the dataset (e.g. `OCC`). Supplies occasions when the model uses `kappa` or `block_kappa` declarations.	Variability: random effects, residual error and IOV
`iov_occasion`	`column`, `dose`, `time(t₁, …)`	`column`	Derive the IOV occasion partition from the model instead of a data column.	Variability: random effects, residual error and IOV
`iscale_max`		`10.0`	Maximum proposal scaling factor.	Estimation methods and controlling the fit
`iscale_min`		`0.1`	Minimum proposal scaling factor for adaptive IS. The IS proposal covariance is multiplied by `s²` where `s` is chosen from `[iscale_min, iscale_max]` to maximise per-subject ESS.	Estimation methods and controlling the fit
`max_unconverged_frac`		`0.1`	Fraction of subjects (with at least `min_obs_for_convergence_check` observations) allowed to have unconverged EBEs before the outer optimizer rejects the step (returns OFV = ∞).	Estimation methods and controlling the fit
`maxiter`	integer	`500`	Maximum outer loop iterations. `maxiter = 0` is evaluation only: the engine runs one inner EBE solve at the initial parameters and reports the objective there with no outer optimisation — useful for scoring a fixed parameter set or computing standard errors at known estimates (with `covariance = true`).	Estimation methods and controlling the fit
`min_obs_for_convergence_check`		`2`	Subjects with fewer than this many observations are excluded from the `max_unconverged_frac` check (they still run normally).	Estimation methods and controlling the fit
`mstep_damping`		`0.03`	Exploration-phase cap on the stochastic-approximation step for the numerical θ/σ M-step, the θ-side counterpart of the per-iteration SA step cap SAEM already applies to Ω.	Estimation methods and controlling the fit
`mu_referencing`	`true`, `false`	`true`	Re-centre inner-loop ETA estimates on the current population mean (auto-detected from `[individual_parameters]`).	Variability: random effects, residual error and IOV
`multi_start_seed`		`42`	RNG seed for the multi-start theta perturbations. Independent of `seed` (SAEM) so that changing the SAEM seed does not silently alter which perturbed starting points are used in FOCE multi-start runs.	Estimation methods and controlling the fit
`n_agq`		`1`	Gauss-Hermite nodes per random effect. `1` reproduces the Laplace approximation exactly; odd values are conventional (they keep a node at the mode).	Estimation methods and controlling the fit
`n_convergence`		`250`	Phase 2 iterations (step size = 1/k)	Estimation methods and controlling the fit
`n_exploration`		`150`	Phase 1 iterations (step size = 1)	Estimation methods and controlling the fit
`n_leapfrog`		`0`	Leapfrog steps per HMC proposal (0 = use MH; see below). When > 0, subjects for which HMC is unavailable (ODE model, missing analytical PK path, non-finite Ω, unsupported TV-cov path) fall back to MH using `n_mh_steps` proposals.	Estimation methods and controlling the fit
`n_mh_steps`		`20`	Block Metropolis-Hastings steps per subject per iteration. Also sizes the componentwise decorrelating kernel that prevents block-Ω collapse (`max(2, n_mh_steps / n_eta)` sweeps; multi-η models only — skipped when `n_eta < 2`).	Estimation methods and controlling the fit
`n_starts`		`1`	Number of independent optimization runs. `1` disables multi-start (no behaviour change).	Estimation methods and controlling the fit
`nn_l2`	float ≥ 0	`0.0`	L2 (weight-decay) strength. Adds `nn_l2 · Σ wᵢ²` over the network’s weight matrices only (bias terms are left free).	Experimental features
`nn_smooth`	float ≥ 0	`0.0`	Smoothness (curvature) strength. Penalizes the finite-difference 2nd derivative `C = f(x+h) − 2·f(x) + f(x−h)` of each output along every input’s marginal partial-dependence curve, i.e. `nn_smooth · Σ C²`.	Experimental features
`npde_nsim`	integer ≥ 0	`0`	Number of Monte-Carlo replicates per subject used to compute the simulation-based NPDE/NPD diagnostics after the fit.	Diagnosing the model
`npde_seed`	integer	—	RNG seed for the NPDE/NPD simulation, for reproducible diagnostics.	Diagnosing the model
`ode_abstol`	float	`1e-6`	RK45 ODE solver absolute tolerance (companion to `ode_reltol`).	Structural models: analytical and ODE
`ode_auto_switch`	`true`, `false`	`true`	Let `ode_method = auto` change stepper inside an integration segment, not only at its start.	Structural models: analytical and ODE
`ode_max_steps`	integer	`10000`	Max solver steps per integration segment. ODE models and the absorption ODE twin.	Structural models: analytical and ODE
`ode_method`	`auto`, `rk45`, `vern7`, `rosenbrock23`, `rodas4`, `rodas5p`	`auto`	Which stepper integrates the `[odes]` block. Two axes: `vern7` (Verner 7(6)) buys order, for fits that are accuracy-limited at tight tolerances — not a blanket upgrade, since it is slower than `rk45` at default tolerances.	Structural models: analytical and ODE
`ode_reltol`	float	`1e-4`	RK45 ODE solver relative tolerance. Applies to ODE models, and to the auto-generated ODE twin that closed-form transit / inverse-Gaussian absorption models (`one_cpt_transit`, `two_cpt_transit`, `one_cpt_ig`, `two_cpt_ig`) use to serve IOV, time-varying-covariate, and `TIME`-switch subjects (#719/#814) — so a call-time override reaches those rerouted fits too; otherwise ignored for analytical PK.	Structural models: analytical and ODE
`ode_stiff_abort_after`	integer ≥ 0, or `off`	`off`	Abandon an integration segment once this many of its steps have clamped at the minimum step size, instead of grinding on to `ode_max_steps`.	Structural models: analytical and ODE
`omega_burnin`		`20`	Initial exploration iterations during which Ω (and Ω_IOV) are held at their starting values while the MH chain warms up.	Estimation methods and controlling the fit
`optimizer`	`auto`, `bobyqa`, `slsqp`, `nlopt_lbfgs`, `mma`, `trust_region` (plus deprecated aliases `bfgs`/`lbfgs` → `nlopt_lbfgs`)	`auto`	Outer-loop optimisation algorithm. Applies to `method = foce` / `focei` and to the FOCEI polish phase of `method = gn_hybrid`.	Estimation methods and controlling the fit
`optimizer_trace`	`true`, `false`	`false`	Write a per-iteration CSV to `/tmp/ferx_trace_<pid>_<ts>.csv`.	Estimation methods and controlling the fit
`outer_ftol`	float	auto	Relative objective tolerance: the relative OFV change below which the optimizer stops.	Estimation methods and controlling the fit
`outer_xtol`	float	`1e-4`	Relative step tolerance: how far the optimizer’s step must shrink before it declares success.	Estimation methods and controlling the fit
`parameter_scaling`	`auto`, `none`, `abs`, `rescale2`	`auto`	Parameter-scaling strategy for the outer optimizer (supersedes `scale_params` when non-`none`).	Variability: random effects, residual error and IOV
`reconverge_gradient_interval`	integer ≥ 0	`0`	How often to re-solve each subject’s inner EBE loop during the population gradient instead of holding the EBEs (η̂) and FOCE Hessian fixed.	Estimation methods and controlling the fit
`report_final_gradient`	`true`, `false`	`true`	Compute a finite-difference gradient at the reported estimates when the optimizer supplied none of its own, and report it as `final_gradient` with `final_gradient_source = "finite_difference"`.	Estimation methods and controlling the fit
`scale_params`		`false`	Legacy boolean alias for `parameter_scaling = abs`. Divide each packed (log/Cholesky) coordinate by its initial magnitude before passing it to the optimizer.	Variability: random effects, residual error and IOV
`seed`		`12345`	RNG seed for reproducibility	Estimation methods and controlling the fit
`sir_df`		`5.0`	Degrees of freedom for the Student-t proposal; higher values approach a normal proposal	Parameter uncertainty
`sir_keep_samples`		`false`	Retain resampled parameter vectors for `simulate_with_uncertainty()`	Parameter uncertainty
`sir_resamples`		`250`	Number of resampled vectors (m)	Parameter uncertainty
`sir_samples`		`1000`	Number of proposal samples (M)	Parameter uncertainty
`sir_seed`		`12345`	RNG seed for reproducibility	Parameter uncertainty
`stagnation_guard`		`true`	Short-circuit the NLopt-based outer optimizers once recent evals show no OFV improvement above 1e-3 over a window of `3*(n+1).max(50)` evals.	Estimation methods and controlling the fit
`start_sigma`		`0.3`	Log-space perturbation applied to initial theta values for starts 1..n.	Estimation methods and controlling the fit
`steihaug_max_iters`	integer	adaptive	Max CG iterations for the Steihaug subproblem (only used when `optimizer = trust_region`).	Estimation methods and controlling the fit
`vi_avg_last`		final 25%	Polyak averaging window, in iterations — the reported estimate is the mean over the final window rather than the last iterate.	Estimation methods and controlling the fit
`vi_eta_grad`		`auto`	`auto` uses the analytic `Dual2` η-gradient where available and central finite differences where not; `analytic` and `fd` pin one route.	Estimation methods and controlling the fit
`vi_family`		`full_rank`	`full_rank` fits a full posterior covariance per subject; `mean_field` fits a diagonal one — `O(d)` rather than `O(d²)`.	Estimation methods and controlling the fit
`vi_final_ofv`		`none`	`none` leaves `ofv` as `NaN`, because the ELBO is a lower bound and is not comparable with a FOCEI OFV; `laplace` computes a comparable one.	Estimation methods and controlling the fit
`vi_grad_clip`		`1e4`	Global-L2 gradient clip applied to every Adam step; `0` disables it.	Estimation methods and controlling the fit
`vi_iters`		`25000`	Adam ceiling, not a fixed budget — the run stops early once the objective and the estimates have both settled.	Estimation methods and controlling the fit
`vi_kl`		`analytic`	How the `KL(q ‖ N(0, Ω))` half is evaluated — `analytic` in closed form (only the data term is sampled) or `mc`.	Estimation methods and controlling the fit
`vi_lr`		`0.02`	Adam learning rate. Lower it if the trace oscillates, or if σ settles above a FOCEI/AGQ fit of the same data.	Estimation methods and controlling the fit
`vi_mc_samples`		`32`	Monte-Carlo draws per subject per iteration. This sets the noise floor the settling test measures against, so it decides when the run stops as well as how noisy each step is; too few draws can settle early at the wrong answer.	Estimation methods and controlling the fit
`vi_omega_update`		`closed_form`	`closed_form` sets Ω to its exact ELBO maximizer each iteration, keeping it out of the stochastic optimization; `adam` learns it with everything else.	Estimation methods and controlling the fit
`vi_seed`		`20240704`	Seed for the common random numbers. See Reproducibility.	Estimation methods and controlling the fit
`vi_sigma_update`		`closed_form`	As `vi_omega_update`, for the residual error σ. See How σ is updated.	Estimation methods and controlling the fit

Model file blocks

The blocks a .ferx model file can contain. Required and Purpose are the ferx-core block overview at the pinned engine, and are blank for a block that overview does not list; the chapter column covers those. Block names are closed-world, so a header outside this table is an error.

Block	Required	Purpose	Chapter	ferx-core page
`[adaptive_dosing]`	No	Feedback / TDM dose controller for `simulate()`	Adaptive dosing and TDM strategies	adaptive-dosing
`[binary_model]`	No	Binary / logistic endpoint (needs `--features survival`)	Binary endpoints	categorical
`[covariate_model]`			Covariate modeling	covariate-model
`[covariate_nn NAME]`	No	Neural-network covariate model (needs `--features nn`)	Experimental features	covariate-nn
`[covariates]`	No	Declare which data columns are covariates (and how they interpolate)	Covariate modeling	covariates
`[data]`	No	Point at the dataset CSV; overridden by an explicit CLI/R path	Preparing and checking the analysis dataset	data
`[data_selection]`	No	Row/subject filters applied to the dataset	Preparing and checking the analysis dataset	data-selection
`[derived]`	No	Post-hoc derived quantities written to sdtab	Tables and figures	derived
`[diffusion]`	No	SDE diffusion terms	Experimental features	diffusion
`[error_model]`	Yes	Define the residual error model	Writing and managing model files	error-model
`[event_model]`	No	Time-to-event endpoint (needs `--features survival`)	Time-to-event models	event-model
`[fit_options]`	No	Configure estimation method and optimizer	Writing and managing model files	fit-options
`[individual_parameters]`	Yes	Map population parameters to individual PK parameters	Writing and managing model files	individual-parameters
`[initial_conditions]`	No	Non-zero compartment initial amounts	Structural models: analytical and ODE	initial-conditions
`[markov_model]`	No	Continuous-time Markov endpoint (needs `--features markov`)	Function, option and example index	markov-model
`[mixture]`	No	Mixture (subpopulation) model	Variability: random effects, residual error and IOV	mixture
`[odes]`	If ODE	ODE right-hand-side equations	Structural models: analytical and ODE	ode-models
`[output]`	No	Extra sdtab output columns	Tables and figures	output
`[parameters]`	Yes	Define theta, omega, and sigma parameters	Writing and managing model files	parameters
`[priors]`			Experimental features	priors
`[scaling]`	No	Compartment scaling / observation readouts	Structural models: analytical and ODE	scaling
`[simulation]`	No	Define a simulation trial design	Simulating scenarios	simulation
`[structural_model]`	Yes	Specify the PK model (analytical or ODE)	Writing and managing model files	structural-model

Data columns

Columns the ferx data format recognises. Type and Meaning are taken from the ferx-core data format page at the pinned engine, and are blank for the columns that page does not list; the chapter column covers those.

Column	Type	Meaning	Chapter
`ADDL`			Preparing and checking the analysis dataset
`AMT`	numeric	Dose amount (for `EVID=1`/`EVID=4`; also the dose-inference signal when `EVID` is absent)	Preparing and checking the analysis dataset
`CENS`	integer	Censoring flag. `1` = below LLOQ and `DV` carries the LLOQ; `-1` = above ULOQ and `DV` carries the ULOQ; `0` = quantified.	Preparing and checking the analysis dataset
`CMT`	integer	Compartment number (1-indexed)	Preparing and checking the analysis dataset
`DV`	numeric	Dependent variable (observed concentration)	Preparing and checking the analysis dataset
`EVID`	integer	Event ID: 0 = observation, 1 = dose, 2 = other event, 3 = system reset, 4 = reset + dose.	Preparing and checking the analysis dataset
`FREMTYPE`			Covariate modeling
`ID`	string/numeric	Subject identifier	Preparing and checking the analysis dataset
`II`	numeric	Interdose interval for repeated dosing	Preparing and checking the analysis dataset
`MDV`	integer	Missing DV flag. 1 = DV should be ignored (row excluded from the likelihood)	Preparing and checking the analysis dataset
`RATE`	numeric	Infusion rate. `0` = bolus, `>0` = constant-rate infusion (duration = `AMT/RATE`).	Preparing and checking the analysis dataset
`SS`	integer	Steady-state flag. 1 = assume steady state	Preparing and checking the analysis dataset
`TENTRY`			Time-to-event models
`TIME`	numeric	Observation/event time on the data clock	Preparing and checking the analysis dataset

Bundled examples

Every example is available with ferx_example() and its name, which returns the paths of its model and data (and $search for a search configuration).

Example	Chapter
`adaptive_tdm`	Adaptive dosing and TDM strategies
`adaptive_vanco_loading`	Adaptive dosing and TDM strategies
`binary_logistic`	Binary endpoints
`bioavailability`	Absorption and bioavailability
`bioavailability_ode`	Absorption and bioavailability
`biphasic_igd_absorption`	Absorption and bioavailability
`emax_pkpd`	PK/PD and multiple endpoints
`emax_timecourse`	PK/PD and multiple endpoints
`igd_inverse_gaussian`	Absorption and bioavailability
`infusion_absorption`	Dosing regimens and exposure metrics
`mixed_absorption`	Absorption and bioavailability
`mm_multistart`	Estimation methods and controlling the fit
`mm_oral`	Structural models: analytical and ODE
`one_cpt_ig`	Absorption and bioavailability
`one_cpt_iv`	Structural models: analytical and ODE
`one_cpt_iv_ode`	Structural models: analytical and ODE
`one_cpt_iv_pooled`	Structural models: analytical and ODE
`one_cpt_transit`	Absorption and bioavailability
`one_cpt_transit_iov`	Variability: random effects, residual error and IOV
`parallel_absorption`	Absorption and bioavailability
`per_route_lag_absorption`	Absorption and bioavailability
`pktte_joint`	Time-to-event models
`sequential_absorption`	Absorption and bioavailability
`ss_absorption`	Dosing regimens and exposure metrics
`three_cpt_iv`	Structural models: analytical and ODE
`three_cpt_iv_ode`	Structural models: analytical and ODE
`three_cpt_oral`	Structural models: analytical and ODE
`three_cpt_oral_ode`	Structural models: analytical and ODE
`transit_2cpt`	Absorption and bioavailability
`transit_savic`	Absorption and bioavailability
`tte_competing_risks`	Time-to-event models
`tte_exponential`	Time-to-event models
`tte_gompertz`	Time-to-event models
`tte_weibull`	Time-to-event models
`two_cpt_ig`	Absorption and bioavailability
`two_cpt_iv`	Structural models: analytical and ODE
`two_cpt_iv_ode`	Structural models: analytical and ODE
`two_cpt_oral_base`	Writing and managing model files
`two_cpt_oral_cov`	Preparing and checking the analysis dataset
`two_cpt_oral_cov_ode`	Structural models: analytical and ODE
`two_cpt_oral_cov_ode_template`	Structural models: analytical and ODE
`two_cpt_oral_derived`	Tables and figures
`two_cpt_oral_global`	Model development and selection
`two_cpt_transit`	Absorption and bioavailability
`warfarin`	A complete analysis in one chapter
`warfarin_additive_eta`	Variability: random effects, residual error and IOV
`warfarin_addl`	Dosing regimens and exposure metrics
`warfarin_amd`	Model development and selection
`warfarin_block_omega`	Variability: random effects, residual error and IOV
`warfarin_bloq`	Censored observations (BLOQ)
`warfarin_data_selection`	Preparing and checking the analysis dataset
`warfarin_dcm`	Experimental features
`warfarin_derived`	Dosing regimens and exposure metrics
`warfarin_derived_pkpd`	PK/PD and multiple endpoints
`warfarin_if`	Covariate modeling
`warfarin_iov`	Variability: random effects, residual error and IOV
`warfarin_iov_saem`	Variability: random effects, residual error and IOV
`warfarin_logit_f`	Absorption and bioavailability
`warfarin_ltbs`	Variability: random effects, residual error and IOV
`warfarin_ode`	Structural models: analytical and ODE
`warfarin_ode_lagtime`	Absorption and bioavailability
`warfarin_ode_time`	Structural models: analytical and ODE
`warfarin_saem`	Estimation methods and controlling the fit
`warfarin_scaled`	Structural models: analytical and ODE
`warfarin_sde`	Experimental features
`warfarin_ss`	Dosing regimens and exposure metrics
`weibull_absorption`	Absorption and bioavailability
`zero_order_absorption`	Absorption and bioavailability

Search configurations

Example with $search	Chapter
`one_cpt_transit`	Model development and selection
`two_cpt_oral_base`	Model development and selection
`two_cpt_oral_cov`	Model development and selection
`two_cpt_oral_global`	Model development and selection
`warfarin`	Model development and selection
`warfarin_amd`	Model development and selection
`warfarin_block_omega`	Model development and selection
`warfarin_iov`	Model development and selection

Warning categories

Categories of ferx_get_warnings(fit, as_df = TRUE)$category, collected by a fit that ran, with the severity and meaning ferx-core gives them at the pinned engine. Critical means the result is untrustworthy as it stands; Warning means it stands with a caveat; Info implies no action. A model that never ran fails with a check report code instead, listed under Check report codes below. Each chapter’s Warnings you may see here section covers the ones it meets, and the ferx-core warnings page describes them all.

Category	Severity	Meaning	Chapter
`absorption_twin_declined`	Warning	An analytic transit / inverse-Gaussian model’s ODE twin could not be built, so the model keeps no ODE fallback (#1008).	Absorption and bioavailability
`bloq_method`	Warning	BLOQ / M3 censoring caveat.	Censored observations (BLOQ)
`boundary_estimate`	Warning	One or more THETA estimates are pinned to an optimizer bound.	Diagnosing the model
`cancelled`	Info	Run cancelled by the user.	Estimation methods and controlling the fit
`condition_number`	Critical	Ill-conditioned covariance / high condition number.	Diagnosing the model
`convergence`	Critical	Optimizer did not converge, or (`W_NONFINITE_OBJECTIVE`) the objective at the final estimates is `NaN`, infinite, or the clamped divergence sentinel.	Initial estimates and a first fit
`covariance_failed`	Critical	Covariance step failed — no standard errors.	Parameter uncertainty
`covariance_regularized`	Warning	Covariance step succeeded but was regularized / degraded.	Parameter uncertainty
`covariance_step`	Info	Covariance-step informational note (e.g. evaluation cost).	Parameter uncertainty
`data_quality`	Warning	Dataset issue (missing DV, ADDL/II, non-positive DV, …).	Preparing and checking the analysis dataset
`dw_autocorrelation`	Warning	IWRES autocorrelation (Durbin–Watson out of range).	Diagnosing the model
`ebe_start_dependent`	Warning	(`W_EBE_START_DEPENDENT`) The empirical Bayes estimates at the final parameters depend on where the inner loop starts: re-solving them cold scores materially worse than the EBEs the optimizer minimised against, or returns a non-finite objective (#833).	Estimation methods and controlling the fit
`eps_shrinkage`	Warning	Residual (EPS) shrinkage high / negative.	Diagnosing the model
`eta_normality`	Warning	ETA departs from normality.	Diagnosing the model
`eta_shrinkage`	Warning	One or more ETA shrinkages exceed ~30%.	Diagnosing the model
`experimental`	Warning	An experimental feature (SDE, neural-network) was used.	Experimental features
`flat_parameter`	Warning	A non-fixed THETA has an ≈ 0 outer gradient at the initial estimate — it never reaches the objective (unmapped, or dropped from the structural / scaling model).	Estimation methods and controlling the fit
`flip_flop`	Warning	A transit / inverse-Gaussian absorption closed form entered the flip-flop regime (disposition rate ≥ the tilting abscissa).	Absorption and bioavailability
`general`	Warning	Unrecognised message (fallback bucket).	Initial estimates and a first fit
`gradient_fallback`	Info	A gradient / sampler fallback was taken.	Estimation methods and controlling the fit
`high_correlation`	Warning	One or more THETA (fixed-effect) pairs have \|correlation\| ≥ 0.95.	Diagnosing the model
`importance_sampling`	Warning	ESS = 0 / proposal collapse.	Estimation methods and controlling the fit
`inflated_rse`	Warning	One or more THETA estimates have RSE > ~50%.	Diagnosing the model
`init_outside_bounds`	Warning	An initial estimate packs outside one of ferx’s internal rails (the hidden `1e9` THETA cap, the OMEGA `±6` / off-diagonal `±10` guards, the SIGMA `[-8, 5]` guard) and was clamped there before the first objective evaluation.	Estimation methods and controlling the fit
`mu_referencing`	Warning / Info	Missing or partial mu-referencing.	Estimation methods and controlling the fit
`multi_start`	Info	Multi-start note.	Estimation methods and controlling the fit
`ode_solver`	Warning / Info	The ODE solver’s own health at the final estimates (#1080). Warning when steps clamped at the minimum step size (a stability-limited segment whose un-integrated tail is freeze-padded), when a segment stopped before its requested end, when `ode_stiff_abort_after` cut segments short, or when `ode_method = auto` discarded an escalation and re-solved it explicitly — including when only its analytic derivatives overflowed, its values staying finite — and when that re-solve failed too.	Structural models: analytical and ODE
`omega_structure`	Warning	Mixed lognormal / additive block in omega.	Variability: random effects, residual error and IOV
`optimizer_config`	Warning / Info	`global_search` config note or failure.	Estimation methods and controlling the fit
`optimizer_health`	Warning	Trust-radius collapse / degeneracy.	Estimation methods and controlling the fit
`parameter_at_runaway_guard`	Critical / Warning	A free coordinate is pinned to a hidden optimizer guard (implicit THETA cap, OMEGA / SIGMA rail), with a `verdict` the side does not decide.	Estimation methods and controlling the fit
`simulation`	Warning	A simulated subject was handled specially — a degenerate hazard draw (censored, no event) or an over-large recurrent-event stream (skipped/truncated).	Time-to-event models
`sir`	Warning	SIR failed or was requested without a covariance.	Parameter uncertainty
`stalled_at_init`	Warning	The fit never left its initial estimates — no free THETA, OMEGA or SIGMA coordinate moved, so the reported objective is the objective of the initial values and carries no information about the model.	Estimation methods and controlling the fit
`threads`	Info	Thread-count efficiency note.	Estimation methods and controlling the fit
`vi_bad_basin`	Critical	VI’s final ELBO tightness check found that a flat objective is an unusable variational bound, not convergence.	Estimation methods and controlling the fit

Check report codes

Stable identifiers from the model file parser and the pre-fit checks: E_* stops the model from running, W_* is a check-time note that does not. ferx_model_validate() returns them in $diagnostics, with the severity, block and line (Writing and managing model files). A refused ferx_fit() raises a condition of class ferx_engine_error that carries the same identifier as $code, with $block, $line and $suggestion where the engine knows them (NA otherwise), and appends the code to the message in square brackets, so a script can branch on it with tryCatch(..., ferx_engine_error = function(e) e$code). These are a different channel from the Warning categories above, which a completed fit collects in fit$warnings. The 49 codes below (31 errors, 18 warnings) are the check report of the pinned engine, one sentence each; the ferx-core check report page gives the full text.

Code	Severity	Meaning
`E_BLOCK_FEATURE_DISABLED`	error	A recognised block needs a cargo feature this binary was not built with — `[event_model]` / `[binary_model]` need `--features survival`, `[markov_model]` needs `--features markov`.
`E_BLOCK_INSTANCE_NAME`	error	A `[block NAME]` instance name is given for a block that takes none (e.g. `[fit_options DOSE]`), or omitted for one that requires it (`[covariate_nn NAME]`).
`E_BLOCK_VARIANCE_ONLY`	error	A scale tag (`(sd)`, `(variance)` or `(var)`) on a `block_omega`, `block_sigma` or `block_kappa` declaration.
`E_DATA`	error	The `--data` file could not be read or parsed.
`E_DEPRECATED_BLOCK`	error	A `[block]` that was ferx syntax and is no longer read — currently `[initial_values]`, whose contents moved inline into `[parameters]`.
`E_DERIVED_NAME_CONFLICT`	error	A `[derived]` name clashes with a built-in sdtab column, theta, eta, or individual-parameter name.
`E_DOSE_ATTR_DOUBLE_USE`	error	A dose attribute the engine applies at the dose event is also read by the model, so its value is applied twice.
`E_DOSE_ATTR_NONFINITE`	error	A dose attribute the engine applies at the dose event — lag time (`lagtime` / `ALAG{n}`), bioavailability (`F` / `F{n}`), or a modeled infusion duration `D{n}` / rate `R{n}` behind a coded `RATE` — is `NaN` or infinite at typical values for some subject.
`E_ENDPOINT_NO_RECORDS`	error	A declared non-Gaussian endpoint has no row routed to it anywhere in the data — typically the `CMT` column is missing, so every row reads as CMT 1.
`E_ENDPOINT_UNROUTED`	error	A compartment the model declares as a non-Gaussian endpoint (`[event_model]`, `[binary_model]`, Markov) carries Gaussian observations: the population was read without endpoint routing, so the endpoint’s rows would be scored as concentrations and its likelihood would contribute nothing.
`E_ETA_NOT_DECLARED`	error	An unresolved identifier reads as a random effect (`ETA…` / `KAPPA…`) but no `omega` / `kappa` declaration defines it, and the data carries no such column.
`E_IMP_CHAIN`	error	`imp` is mis-placed in a method chain — repeated, or (with `imp_eval_only = true`) not the terminal stage.
`E_INIT_BOUNDS_INVERTED`	error	A coordinate whose packed box is empty, so no start can be placed in it at all.
`E_IOV_MISSING_OCC`	error	Model declares kappa (IOV) parameters but no occasion labels were found in the dataset.
`E_METHOD_NO_RANDOM_EFFECTS`	error	`method = imp`, `impmap` or `bayes` (anywhere in a method chain) on a model with no random effects (`n_eta = 0`).
`E_MISSING_BLOCK`	error	A required `[block]` is absent. `block` names which one.
`E_MISSING_COVARIATE`	error	The model references a covariate not present in the data (case-sensitive).
`E_NN_FEATURE_DISABLED`	error	A `[covariate_nn]` block requires building with `--features nn`.
`E_NONFINITE_DV`	error	A subject carries a non-finite (`NaN`/`Inf`) observation. Usually a population read by `read_population_for_simulation()` — a simulation input — passed to `fit()`.
`E_OMEGA_INIT_AT_RAIL`	error	A free `omega` / `kappa` / `[mixture] omega(k)` variance whose initial value packs onto the optimizer’s `-6` lower rail — every variance ≤ 6.1e-6, and `~ 0.0` in particular (a declared zero is regularised to 1e-8, packing to `ln(L) = -9.21`).
`E_OPTIMIZER_AGQ`	error	`optimizer = trust_region` used with a quadrature stage (`method = laplace`, or `method = focei` with `n_agq > 1`).
`E_OPTIMIZER_IOV`	error	`optimizer = trust_region` used with an IOV model (`n_kappa > 0`).
`E_OUTPUT_UNKNOWN_COLUMN`	error	A name in `[output]` is not recognised as a covariate, individual parameter, or derived expression.
`E_PARSE`	error	The model file failed to parse (catch-all for parser errors).
`E_PER_CMT_ERROR_MODEL`	error	An observed compartment lacks a per-CMT `[error_model]` entry.
`E_PER_CMT_SCALING`	error	An observed compartment lacks a per-CMT scaling entry.
`E_SAEM_NO_RANDOM_EFFECTS`	error	`method = saem` (anywhere in a method chain) on a model with no random effects (`n_eta = 0`).
`E_SDE_INCOMPATIBLE`	error	An SDE (`[diffusion]`) model used with an incompatible method (`saem`, `gn`, `gn_hybrid`) or gradient setting.
`E_SIGMA_ORDER_MISMATCH`	error	A single-endpoint `[error_model]` names its sigmas in an order other than the `[parameters]` declaration order.
`E_THETA_INIT_OUTSIDE_BOUNDS`	error	A `theta` whose initial value is strictly outside its own declared range. The optimizer clamps the start into its box, so `theta TVCL(0.05, 0.1, 10.0)` fits from `0.1` — a factor of two — silently, on every run.
`E_UNKNOWN_BLOCK`	error	A `[block]` header is not a recognised block name. The message lists every offending header with its line, the valid set this build accepts, and a did-you-mean when there is a near match; `suggestion` carries the near match on its own.
`W_ABSORPTION_TWIN_DECLINED`	warning	An analytic transit / inverse-Gaussian model’s ODE twin could not be built, so the model stays closed-form with no ODE fallback.
`W_ADDL_MISSING_II`	warning	ADDL > 0 on a dose row but II is zero or missing; additional doses were not expanded.
`W_DERIVED_COVARIATE_SHADOW`	warning	A `[derived]` name shadows a covariate (allowed but may be confusing).
`W_DERIVED_STEP_IGNORED`	warning	`step=` given for a DV-based integral (ignored; DV integrals always use observation times).
`W_DESIGN_DV`	warning	Simulation reader only: one or more `EVID=0` rows had a missing `DV` and were kept as design points to simulate at, so the dataset contributes more rows than a fit of the same file would.
`W_ETA_NOT_DECLARED`	warning	As `E_ETA_NOT_DECLARED`, but reported without `--data`, where ferx cannot yet know whether the dataset supplies the column.
`W_EXPERIMENTAL_NN`	warning	The model uses neural-network components (`[covariate_nn]`), an experimental feature.
`W_EXPERIMENTAL_SDE`	warning	The model uses stochastic differential equations (`[diffusion]`), an experimental feature.
`W_GN_NO_RANDOM_EFFECTS`	warning	`method = gn` as the last estimating stage on a model with no random effects (`n_eta = 0`).
`W_INIT_OUTSIDE_BOUNDS`	warning	An initial estimate strictly outside one of ferx’s internal rails rather than a declared bound: the hidden `1e9` θ cap, the Ω `±6` (variance ≤ 1.6e5) / off-diagonal `±10` guards, or the Σ `[-8, 5]` guard.
`W_IOV_OCC_MISSING`	warning	Some rows in the IOV occasion column had missing or unparseable values; those rows were assigned occasion=0.
`W_MISSING_DV`	warning	One or more `EVID=0` rows had a missing `DV` (`.`/`NA`/blank) but were not marked `MDV=1`; they were skipped rather than scored as `DV=0`.
`W_NEGATIVE_LAGTIME`	warning	Lag time is negative at the initial typical-value point.
`W_OUTPUT_DUPLICATE`	warning	A name in `[output]` is already written to sdtab automatically (e.g. `TAFD`, `TAD`, an eta name).
`W_SDE_RESET`	warning	EVID=3/4 resets under an SDE `[diffusion]` model are not honoured.
`W_STEADY_STATE_ABSOLUTE_TIME`	warning	An `SS=1` dose (`II > 0`) on a model whose `[odes]` right-hand side reads an absolute clock — `TAFD`, `T`/`t`, or the bare `TIME` built-in.
`W_STEADY_STATE_II`	warning	SS=1 doses with missing / non-positive `II` (treated as non-SS).
`W_STEADY_STATE_INFUSION`	warning	SS=1 infusion whose length after bioavailability exceeds `II` (overlapping pulses; SS skipped) — the run-in’s own test (#1281).

Fit object slots

Elements of the list ferx_fit() returns, for a fit of the warfarin example. Some slots are NULL unless the corresponding step or model feature is used. Descriptions are the first sentence of the entry in ?ferx_fit, where it has one.

Slot	Description
`aic`	Akaike Information Criterion
`bayes`
`bic`	Bayesian Information Criterion
`bic_inputs`	Named list of the free-parameter tally by Delattre class - n_obs, theta_random, theta_fixed, omega, kappa, sigma, sigma_random - which ferx_bic turns into the mixed / IIV / random BIC variants.
`bloq_method_label`	Character string describing the LOQ-censoring handling method used (“m3”, “drop”, etc.).
`call_settings`	Named list of the effective settings passed to Rust, with values typed back to their natural R types (logical, numeric, or character).
`cond_dist`
`condition_number`	Ratio of the largest to smallest eigenvalue of the correlation matrix of estimated parameters.
`converged`	Logical; did the optimizer converge
`cor_matrix`	Correlation matrix derived from cov_matrix, same dimnames. A correlation close to between two parameters flags a structural identifiability problem in the model.
`cov_matrix`	Full parameter covariance matrix as a named numeric matrix (params × params). Row/column names use declared variable names (“TVCL”, “ETA_CL”, “EPS_PROP”); fallback is “OMEGA(1,1)” / “SIGMA(1)”.
`covariance_n_evals_estimated`	Estimated number of OFV calls required for the covariance step, computed before the step runs on large models.
`covariance_status`	String: “computed”, “failed”, “not_requested”, or “sir_fallback” (the finite-difference Hessian was not positive definite and covariance_fallback = “sir” produced SIR-based intervals).
`covariate_names`	Character vector of non-standard column names present in the dataset (beyond ID, TIME, DV, EVID, AMT, CMT, RATE, MDV, II, SS, CENS, OCC).
`covariate_types`	Named character vector mapping each declared covariate to its type, “continuous” or “categorical” (the type from the [covariates] block).
`covtab`	Data frame echoing the declared covariate columns, present only when the model has a [covariates] block.
`data_hash`
`data_name`	Dataset name, derived as the basename of the data path with its extension stripped.
`data_path`
`dw_statistic`	Pooled Durbin-Watson statistic for IWRES within subjects. 2.0 = no autocorrelation; < 1.5 = positive autocorrelation (possible missing dynamics); > 2.5 = negative autocorrelation (possible over-parameterisation).
`ebe_convergence_warnings`	Number of outer iterations in which too many EBEs were unconverged (step was rejected by the guard).
`ebe_etas`	Data frame with one row per subject containing the BSV empirical Bayes estimates: ID plus one column per eta named after the model’s eta declarations (e.g. ETA_CL, ETA_V).
`eigenvalues`	Numeric vector of eigenvalues of the correlation matrix of estimated (non-fixed) parameters, sorted descending.
`estimate_near_boundary`	TRUE when a theta is pinned to a declared bound - the predicate ferx_bootstrap applies as skip_estimate_near_boundary.
`estimates`	Tidy data frame of all estimated parameters (theta, omega diagonal, sigma, and - for IOV models - kappa diagonal), with columns param, transform, estimate, se, rse_pct, lower_95, upper_95, estimate_natural, lower_95_natural, upper_95_natural, init_as_sd, weight.
`eta_linked_theta`
`eta_log_transformed`	Logical vector of length n_eta; TRUE when the eta is lognormally parameterised (THETA * exp(ETA)), FALSE for additive or unknown parameterisations.
`eta_names`	Character vector of ETA parameter names as declared in the model file (e.g. “ETA_CL”, “ETA_V”).
`eta_normality`	Data frame with Shapiro-Wilk normality test for each ETA: columns eta, W, p_val, flag.
`eta_param_types`
`exclusions`	Named list summarising records excluded by [data_selection] rules (model file and/or ignore/ accept/ignore_ids arguments).
`ferx_version`	ferx-core library version string.
`final_gradient`	Numeric vector containing the gradient of the objective function at the best-OFV parameter point (packed space: log-theta, Cholesky-omega, log-sigma).
`gradient`	The inner-loop gradient method as requested by the caller (usually “auto” or “fd”).
`gradient_method_inner`	Engine label for the inner-loop gradient method. Prefer gradient_used for display.
`gradient_method_outer`	Engine label for the population-level optimizer’s gradient method.
`gradient_used`	The inner-loop gradient method the engine actually used: “analytic”, “fd”, or “N/A” (derivative-free / sampling step).
`imp_seed`	Numeric scalar (or NULL) giving the importance sampling seed. NULL when IS was not run.
`impmap_trace`
`importance_sampling`	Importance-sampling marginal log-likelihood diagnostics, populated only when the method chain ends with an “imp” stage (NULL otherwise).
`individual_estimates`	Data frame with one row per subject containing each subject’s individual parameter values: ID plus one column per parameter declared in the [individual_parameters] block (e.g. CL, V, KA).
`inits_from_nca`	Character string giving the NCA initialisation method used (“nca”, “nca_sweep”, or “nca_ebe”), or NULL when model-file initial values were used directly.
`input_columns`	Includes both standard columns (ID, TIME, DV, etc.) and any covariate columns. Empty for in-memory fits that never read a file and absent after a ferx_save_fit/ferx_load_fit round-trip on older bundles.
`iwres_lag1_r`	Pooled lag-1 Pearson correlation of IWRES within subjects. Positive values indicate under-fitting, negative values indicate over-fitting.
`kappa_init_as_sd`	Logical vector, one entry per IOV kappa parameter. Same semantics as omega_init_as_sd for inter-occasion variability.
`kappa_is_diagonal`	The same flag for the IOV (KAPPA) block; NA when the model declares no IOV.
`left_init`	The outer optimizer’s own verdict on whether it left the initial estimates: TRUE, FALSE, or NA when the method recorded none.
`max_abs_correlation`	Largest absolute off-diagonal of the covariance matrix in correlation form, on the natural theta / OMEGA / SIGMA scale.
`max_unconverged_subjects`	Worst-case number of unconverged subjects in a single outer iteration.
`method`	Estimation method used
`method_chain`
`model_file_settings`	Named list of raw key = value pairs from the model file’s [fit_options] block, as character strings (before Rust type coercion).
`model_hash`
`model_name`	Model name declared in the .ferx file by a top-level model <name> line (outside any [block], e.g. model warfarin_pk); the name is a single word.
`model_path`
`model_source`
`model_structure`	Named list returned by the Rust engine, derived from the parsed CompiledModel so it reflects exactly what ferx-core ran.
`model_text`	Verbatim content of the .ferx model file as a character string. NULL for in-memory fits that were never associated with a file, or for fits produced by older engine versions.
`multi_start_seed`	Numeric scalar (or NULL) giving the random seed used for multi-start parameter perturbation.
`n_iterations`	Number of outer iterations run
`n_obs`
`n_parameters`
`n_starts`	Integer. Number of multi-start runs attempted. 1 for single-start fits.
`n_subjects`
`n_threads_used`	Integer. Actual number of parallel threads used by the engine during fitting.
`neural_networks`	Named list of sub-lists, one per [covariate_nn] block declared in the model file.
`nlopt_missing_algorithms`	Character vector of NLopt algorithm names that are not available in the current build (empty on most platforms).
`obs_time_range`	Length-2 numeric vector c(min_time, max_time) across all observation records, or NULL when there are no observations.
`ofv`	Objective function value (-2 log-likelihood). The penalized objective when the model declares a prior(…) on any parameter, i.e. ofv_data + ofv_prior.
`ofv_data`	The data half of ofv - the -2 log-likelihood the AIC and BIC are computed from, and the value to compare against an unpriored fit of the same model.
`ofv_prior`	The prior half of ofv: the penalty summed over the priored coordinates. Exactly 0 when no prior is declared.
`omega`	Between-subject variability covariance matrix. Row and column names are the declared ETA names (e.g. “ETA_CL”); fallback is “OMEGA(1,1)” when names are absent.
`omega_init`	Initial omega variance matrix, same layout as omega. A zero matrix for in-memory fits and older bundles.
`omega_init_as_sd`	Logical vector, one entry per BSV eta. TRUE when the corresponding omega line in the model file was annotated with (sd), meaning the initial value was specified as a standard deviation rather than a variance. print.ferx_fit() appends [initial specified as SD] to those rows.
`omega_init_dim`	Number of rows/columns of omega_init.
`omega_is_diagonal`	TRUE when the OMEGA block the fit was estimated under is diagonal, FALSE for a block_omega model, NA when the engine recorded no layout.
`optimizer_label`	Human-readable label for the outer optimizer used (e.g. “LBFGS”, “BOBYQA”). Mirrors the optimizer enum value from ferx-core.
`outer_gtol`	Numeric. Gradient tolerance passed to the outer optimizer.
`outer_maxiter`	Integer. Maximum number of outer-loop iterations passed to the optimizer.
`prior_summary`	Per-parameter prior report, or NULL when the model declares no prior. One row per priored parameter with columns name, prior_value, estimate, shift_in_prior_sds (signed, in prior SDs), penalty (this parameter’s contribution to ofv_prior), family (“lognormal” or “normal”, decided by the parameter’s declared lower bound), and the implied 95% prior interval prior_lower_95 / prior_upper_95.
`saem_mu_ref_m_step_evals_saved`	Number of OFV evaluations saved by mu-referencing in the SAEM M-step. Only populated for SAEM fits with mu_referencing = TRUE; NULL otherwise.
`saem_n_subjects_hmc`	Integer count of subjects whose SAEM E-step proposals were generated by HMC (i.e. n_leapfrog > 0 and the AD build is active).
`saem_seed`	Numeric scalar (or NULL) giving the SAEM stochastic seed. NULL for non-SAEM methods.
`sdtab`	Data frame with ID, TIME, DV, PRED, IPRED, CWRES, IWRES, EBE_OFV, N_OBS; OCC if any subject carries an occasion column; CENS if any rows are LOQ-censored; CMT if the dataset has more than one endpoint (any observation with CMT != 1).
`se_omega`	Standard errors for omega elements. For diagonal omega, a vector of length n_eta (one per variance).
`se_sigma`	Standard errors for sigma (on the SD scale, like sigma itself)
`se_theta`	Standard errors for theta (NULL if covariance step failed)
`shrinkage_eps`	EPS shrinkage: 1 - SD(IWRES). NA when fewer than 2 valid residuals.
`shrinkage_eta`	Numeric vector of ETA shrinkage per random effect (1 - SD(eta_hat_k) / sqrt(omega_kk)).
`sigma`	Residual error parameter estimates on the standard-deviation scale. For a proportional component CV% = sigma * 100; for an additive component the value is in observation units.
`sigma_init`	Numeric vector of initial sigma values, parallel to sigma and sigma_names.
`sigma_init_as_sd`	Logical vector, one entry per sigma component. Same semantics as omega_init_as_sd for residual-error parameters.
`sigma_names`	Names of the sigma parameters as declared in the [parameters] block of the .ferx file.
`sigma_types`	Per-component error type label, parallel to sigma: “proportional” or “additive”.
`sir_resamples`
`sir_resamples_dim`
`sir_resamples_n`
`sir_seed_used`	Numeric scalar (or NULL) giving the SIR resampling seed. NULL when SIR was not run.
`stalled_at_init`	TRUE when no free parameter moved more than 1% of its initial value, when the fit carries nothing to judge by.
`theta`	Named numeric vector of fixed effect estimates
`theta_init`	Numeric vector of initial theta values as supplied to the optimizer, parallel to theta and theta_names.
`theta_transforms`
`total_ebe_fallbacks`	Total Nelder-Mead fallback invocations across all subjects and iterations.
`uses_sde`	Logical; TRUE when the model file contained a [diffusion] block and the fit used the Extended Kalman Filter (EKF) likelihood.
`wall_time_secs`	Total wall-clock time for the fit in seconds.
`warnings`	Character vector of warnings
`warnings_category`
`warnings_message`
`warnings_severity`
`warnings_source_method`
`warnings_structured`	Data frame with columns severity (“critical”, “warning”, or “info”), category (a fixed vocabulary such as “convergence”, “covariance_step”, “dw_autocorrelation”), message, and source_method (the estimation stage that emitted the warning, e.g. “FOCEI”; empty string when not applicable).

Not available from R

Features documented for ferx-core that the pinned ferx-r build cannot run, or runs only partly:

Feature	Status at the pinned build	Where to look
Variational inference as `ferx_fit(method = "vi")`	The `method` argument rejects it; `method = vi` in `[fit_options]` runs	Estimation methods and controlling the fit
`[markov_model]` (continuous-time Markov endpoints)	Rejected at parse time: ferx-r does not build the engine’s `markov` feature	Markov models
`[simulation]` block	Read by the ferx-core command line only; ferx-r’s simulation functions need a dataset	Simulating scenarios
`[dynamics_nn]` (neural-network ODE terms)	Not implemented (design only)	Neural networks
Hand-written `[covariate_model]` relations, repeated time-to-event, fixed-rate infusions into the central compartment	Fit from R, but no bundled ferx-r example	Covariate modeling, Time-to-event models, Dosing regimens and exposure metrics
`[mixture]` models, modelled infusion rate or duration (`RATE = -1`, `-2`)	No bundled example	Variability: random effects, residual error and IOV, data format
VI’s ELBO and per-subject posterior summaries	Not on the R fit object	Variational inference

Feature maturity

The maturity label of each ferx-core feature at the pinned engine (ferx-core d66046e). stable features are well tested across datasets and estimation options; beta features are stable in limited testing; experimental features have been tested on a few examples and may change. See the ferx-core feature maturity page.

Feature	Maturity	ferx-core page
Parameters (theta / omega / sigma, block omega)	stable	Parameters
Inter-occasion variability (IOV)	stable	IOV
Covariates	stable	Covariates
Structural model — analytical PK (1/2/3-cpt)	stable	Structural Model
Lag time	stable	Lagtime
Steady-state doses (SS)	stable	Steady-State Doses
Multiple dosing (ADDL / II)	stable	Multiple Dosing
Error model (additive / proportional / combined)	stable	Error Model
Simulation	stable	Simulation
Individual parameters DSL	beta	Individual Parameters
BLOQ / censored observations (M3)	beta	BLOQ
ODE models (Dormand-Prince RK45)	beta	ODE Models
Stiff ODE steppers (`ode_method` = Rosenbrock23 / Rodas4 / Rodas5P)	beta	Stiff systems
High-order ODE stepper (`ode_method` = Vern7)	experimental	Choosing a method
Automatic stiff-solver selection (`ode_method = auto`, the default)	beta	Letting ferx pick the stepper
Built-in absorption models (transit / inverse-Gaussian / Weibull)	beta	Absorption
Scaling	beta	Scaling
Data selection	beta	Data Selection
Derived columns	beta	Derived Columns
Output columns	beta	Output Columns
Time-to-event endpoints (TTE)	beta	Time-to-Event Endpoints
Stochastic differential equations (SDE)	experimental	SDE
Neural networks (DCM / NODE)	experimental	Neural Networks
Adaptive (feedback) dosing — `[adaptive_dosing]` block / `simulate_adaptive()`	beta	Adaptive Dosing
FOCE / FOCEI	stable	FOCE / FOCEI
Gauss-Newton (BHHH) — `gn`, `gn_hybrid`	beta	Gauss-Newton
SAEM	beta	SAEM
SIR	beta	SIR
Importance sampling (IMP)	beta	Importance Sampling
Outer optimizers (BOBYQA, SLSQP, L-BFGS, MMA, trust-region)	beta	Outer Optimizers
Time-to-event estimation (TTE)	beta	Time-to-Event