In greta: Simple and Scalable Statistical Modelling in R

Logistic regression with error term

This is an example of a simple logistic regression with an extra observation-level error term, to model over-dispersion or clustering in occupancy data from multiple visits.

data

# make fake data
n_env <- 3
n_sites <- 20
n_obs <- 5

# n_sites x n_env matrix of environmental variables
env <- matrix(rnorm(n_sites * n_env), nrow = n_sites) 
# n_sites observations of species presence or absence over n_obs visits
occupancy <- rbinom(n_sites, n_obs, 0.5)

greta code

alpha <- normal(0, 10)
beta <- normal(0, 10, dim = n_env)
error <- normal(0, 10, dim = n_sites)

# logit-linear model with extra variation
linear_predictor <- alpha + env %*% beta + error
p <- ilogit(linear_predictor)

# distribution (likelihood) over observed values
distribution(occupancy) <- binomial(n_obs, p)

greta documentation built on Sept. 8, 2022, 5:10 p.m.