glmmTMB via RTMB

Background

The glmmTMB package fits generalized linear mixed models with extended features such as zero-inflation, flexible dispersion, and many additional correlation structures. It was developed with a user-interface based on lme4 and using the TMB package for efficient and robust computations. The TMB package requires the user (in this case the glmmTMB developer / contributor) to code computations using a superset of C++. However, the C++-like TMB language is no longer needed for these complex calculations because a simpler, more R-like, language RTMB was introduced several years ago and has been gaining popularity. Most courses that were formerly taught in TMB have replaced it with RTMB. Thus, we anticipate that there will be more future glmmTMB developers / contributors available if the computations are written in RTMB.

The glmmTMB package is popular across many research fields, with 13,167 citations on Google Scholar for the R-suggested citation and an additional 870 for other formats of the citation. According to https://cranlogs.r-pkg.org, there are 32K downloads per month. According to https://www.datasciencemeta.com/rpackages, there have been 1,725,397 downloads in total. There were originally 10 developers (of which a handfull are still at least occasionally active) and it has aquired 13 official contributors. At least 2 PhD students have contributed to the package as part of their dissertation work.

Related work

The brms package has most of the same functionality (and more), but it requires a Bayesian flavor which takes much more computation time. Other packages such as lme4, mgcv, and sdmTMB have some of the funtionality, but not everything. sdmTMB was forked from glmmTMB but it may lack some of the distribution families and random effect correlation structures as well as compatibility with some packages for model evaluation and presentation. lme4 lacks zero-inflation and dispersion models. mgcv lacks some of the distribution families and random effect correlation structures.

Details of your coding project

We expect the contributor to fork our package on GitHub and make the following changes as time allows:

• discuss parallelization options with RTMB developer (K Kristensen) to determine feasability
• determine if utils.cpp is still needed (16 lines)
• determine if init.h is still needed (27 lines)
• determine if Makevars is still needed (3 lines)
• convert glmmTMB.cpp from C++ to RTMB code (1386 lines)
• convert contrib.h from C++ to RTMB code (148 lines)
• convert distrib.h from C++ to RTMB code (440 lines)
• rewrite Make file
• rewrite documentation of how users can add functionality vignettes/hacking.rmd

Tests of correctness will be performed using our built-in unit tests that will expect the forked version to give the same results as the current CRAN version.

Speed comparisons of the old and new version will be performed using large data sets that M Brooks has, including spatial and temporal correlation.

If extra time remains, the contributor could make a new version of the selfisher package which is specifically developed for fisheries gear selectivity models. It was created by forking glmmTMB in 2017 and rewriting the model formulas. Since forking, the two packages have diverged immensely. It would be ideal to set up the new selfisher fork in a way that facilitates glmmTMB developments to be easily ported over. Then it should be easier to get and maintianselfisher on CRAN.

Expected impact

This will make it easier for members of the R community to contribute to glmmTMB in the future because the internal computations will be done in a language that is very similar to R. Based on community-initiated developments and discussions, we know that the community likes to add correlation structures, distribution families, and link functions.

Mentors

Contributors, please contact mentors below after completing at least one of the tests below.

• EVALUATING MENTOR: Mollie Brooks molbr@aqua.dtu.dk is the maintiner of R packages glmmTMB and selfisher. She teaches a semester-long course on RTMB each spring.
• Ben Bolker bbolker@gmail.com is an expert in R programming, and has previous GSOC experience with lme4 via the RStats community in 2025. Ben is the most active developer of glmmTMB and has taught RTMB.
• Kasper Kristensen kaskr@dtu.dk is the maintiner of R packages TMB and RTMB.

Tests

Contributors, please do one or more of the following tests before contacting the mentors above.

MENTORS: write several tests that potential contributors can do to demonstrate their capabilities for this particular project. Ask some hard questions that will give you insight about how the contributors write code to solve problems. You'll see that the harder the questions that you ask, the easier it will be for you to choose between the contributors that apply for your project! Please modify the suggestions below to make them specific for your project.

• Easy: something that any useR should be able to do, e.g. download some existing package listed in the Related Work, and run it on some example data.
• Medium: something a bit more complicated. You can encourage contributors to write a script or some functions that show their R coding abilities.
• Hard: Can the contributor write a package with Rd files, tests, and vignettes? If your package interfaces with non-R code, can the contributor write in that other language?

Solutions of tests

Contributors, please post a link to your test results here.

• EXAMPLE CONTRIBUTOR 1 NAME, LINK TO GITHUB PROFILE, LINK TO TEST RESULTS.

1. Name: Siddhesh Akole , GITHUB PROFILE: "Siddd-hrr"

• Easy-test, Medium-test, Hard-test