Fast predictions from glmmTMB models

This repository contains a single function, predict_delta_glmmTMB, which returns predicitons and confidence intervals for model predictions fitted with glmmTMB. Confidence intervals are computed using the Delta method. The function is much faster and uses less memory than ggeffects::ggpredict, but does not perform any checks.

Example

1. Fake data

set.seed(2)
n = 100

# covariates
data = replicate(2, rnorm(n, 0,1))

# parameters
b = c(0.5, 0.2)
sd = 0.2

# response
y = 0 + data %*% b + rnorm(n, sd = sd)
data = data.frame(y = y, data)

2. Fit model

## Warning in checkDepPackageVersion(dep_pkg = "TMB"): Package version inconsistency detected.
## glmmTMB was built with TMB version 1.9.11
## Current TMB version is 1.9.17
## Please re-install glmmTMB from source or restore original 'TMB' package (see '?reinstalling' for more information)

3. Generate predictions

source('glmmTMB_predict_delta.R')

#newdata
newdata <- expand.grid(X1 = seq(-1.5, 1.5, length.out = 100),
                       X2 = c(-1.5, 0, 1.5))


pred <- predict_delta_glmmTMB(m, newdata)
head(pred)

##          X1   X2        fit  conf_low  conf_high conf_level
## 1 -1.500000 -1.5 -1.0013423 -1.095960 -0.9067242       0.95
## 2 -1.469697 -1.5 -0.9863839 -1.080365 -0.8924025       0.95
## 3 -1.439394 -1.5 -0.9714256 -1.064779 -0.8780725       0.95
## 4 -1.409091 -1.5 -0.9564673 -1.049200 -0.8637343       0.95
## 5 -1.378788 -1.5 -0.9415090 -1.033631 -0.8493874       0.95
## 6 -1.348485 -1.5 -0.9265507 -1.018069 -0.8350319       0.95

4. Create plot

library(ggplot2)
(pred_plot <- ggplot(pred, aes(x = X1, y = fit)) +
  geom_line(aes(colour = as.factor(X2))) +
  geom_ribbon(aes(fill = as.factor(X2), ymin = conf_low, ymax = conf_high), alpha = 0.2) + 
  theme_minimal() + 
  theme(aspect.ratio = 1,
        legend.position = 'inside',
        legend.position.inside = c(0.8, 0.2)))

5. Compare to ggeffects::ggpredict

library(ggeffects)
# generate predictions
pred_ggpredict <- data.frame(ggpredict(m, newdata))

(pred_plot_ggeffects <- ggplot(pred_ggpredict, aes(x = x, y = predicted)) +
  geom_line(aes(colour = group)) +
  geom_ribbon(aes(fill = group, ymin = conf.low, ymax = conf.high), alpha = 0.2) + 
  theme_minimal() + 
  theme(aspect.ratio = 1,
        legend.position = 'inside',
        legend.position.inside = c(0.8, 0.2)))