nlmixr2lib

Which version are you on?

This site documents the development version. The articles here cover every model in the development library; the CRAN release contains far fewer, so a readModelDb() call copied from an article will fail with 'name' not in database if you are on CRAN.

	models
CRAN 0.3.2 (released 2026-01-18)	59
development 0.3.2.9000	2912

To get the models these articles describe:

install.packages(
  "nlmixr2lib",
  repos = c("https://nlmixr2.r-universe.dev", getOption("repos"))
)

Then restart R. readModelDb() and modeldb both resolve against the package as loaded, so a reinstall in a running session leaves the old model list in memory and the new models will still appear missing.

Check it worked:

packageVersion("nlmixr2lib")   # 0.3.2.9000
nrow(nlmixr2lib::modeldb)      # 2912

The r-universe repository is needed rather than plain install_github(): the development version requires rxode2 (>= 5.1.8), which is not on CRAN yet, and r-universe serves prebuilt binaries so no compiler is required.

This is a model library for nlmixr2. The package allows a few ways to interact with the model library:

# See all available models
modellib()
# Load the "PK_1cmt" model
modellib(name="PK_1cmt")
# Switch residual error to additive
modellib(name="PK_1cmt", reserr = "addSd")
# Add inter-individual variability on ka and v and switch residual error to
# additive and proportional
modellib(name="PK_1cmt", eta = c("lka", "lv"), reserr = c("addSd", "propSd"))

Modifying models by piping

You may also modify any model from the library (or your own models) with a piping interface. The code below adds inter-individual variability on ka and v and then switches residual error to additive and proportional.

modellib(name="PK_1cmt") |>
  addEta(c("lka", "lv") |>
  addResErr(c("addSd", "propSd"))

Possible extensions

The modellib function is set-up in way that it can be easily extended and used in other applications. A possible extension could be implementation in a shiny app. An app can be created to easily add new models to the model library database (curated?), and directly make these models available for other users. I believe there can be added value in having a base model library that can be easily extended by the community this way.