Parallelize 'glmmTMB' functions

Henrik Bengtsson

+ =

The futurize package allows you to easily turn sequential code into parallel code by piping the sequential code to the futurize() function. Easy!

TL;DR

library(futurize)
plan(multisession)
library(glmmTMB)

m <- glmmTMB(count ~ mined + (1 | site), data = Salamanders, family = nbinom2)
ci <- confint(m, method = "profile") |> futurize()

Introduction

This vignette demonstrates how to parallelize glmmTMB functions such as confint() and profile() through futurize().

The glmmTMB package fits generalized linear mixed models (GLMMs) using Template Model Builder (TMB). Its confint() and profile() functions compute confidence intervals and likelihood profiles for model parameters. When using method = "profile" or method = "uniroot", these computations are performed independently for each parameter, making them candidates for parallelization.

Example: Profile likelihood confidence intervals

The confint() function computes confidence intervals for parameters of a fitted glmmTMB model. Using method = "profile" runs a profile likelihood for each parameter. For example, using the built-in Salamanders dataset to model salamander counts:

library(glmmTMB)

## Fit a negative binomial GLMM
m <- glmmTMB(count ~ mined + (1 | site), data = Salamanders, family = nbinom2)

## Compute profile likelihood confidence intervals
ci <- confint(m, method = "profile")

Here confint() is calculated sequentially. To calculated in parallel, we can pipe to futurize():

library(futurize)
library(glmmTMB)

m <- glmmTMB(count ~ mined + (1 | site), data = Salamanders, family = nbinom2)
ci <- confint(m, method = "profile") |> futurize()

This will distribute the per-parameter profile computations across the available parallel workers, given that we have set up parallel workers, e.g.

plan(multisession)

The built-in multisession backend parallelizes on your local computer and works on all operating systems. There are other parallel backends to choose from, including alternatives to parallelize locally as well as distributed across remote machines, e.g.

plan(future.mirai::mirai_multisession)

and

plan(future.batchtools::batchtools_slurm)

Example: Likelihood profile

The profile() function computes the likelihood profile for each model parameter, which confint() uses internally when method = "profile". We can call it directly and parallelize it using:

library(futurize)
plan(multisession)
library(glmmTMB)

m <- glmmTMB(count ~ mined + (1 | site), data = Salamanders, family = nbinom2)
pr <- profile(m) |> futurize()

Supported Functions

The following glmmTMB functions are supported by futurize():

• confint() for ‘glmmTMB’ (with method = "profile" or method = "uniroot")
• profile() for ‘glmmTMB’