Simulated catch of lobsters per pot from 2000 to 2017, with changing seasonal, depth, and soak-time coverage. The teaching scenario deliberately shifts sampling towards higher-catch months, deeper fishing during the middle years, and longer soak times in the later years. These changes can obscure the known annual effect until the relevant covariates are included. The relationships are specified for demonstration and are not empirical estimates of lobster catch relationships. All values are generated by the deterministic simulator in `data-raw/compact-brms-fixtures.R`; no commercial fishing records are included.
Usage
data(lobsters_per_pot)Format
A data frame with 5,049 simulated pot lifts and five columns:
- lobsters
Number of lobsters caught in one pot.
- year
Factor identifying fishing year, from 2000 to 2017.
- month
Two-digit factor identifying calendar month.
- depth
Fishing depth in metres.
- soak
Pot soak time in hours.
Details
The response follows a negative-binomial distribution with a known log-scale annual effect, seasonal effect, quadratic depth effect, and linear soak-time effect. Deliberately missing year-month cells and uneven sampling make the influence and refitted step examples informative.
The simulation attribute records seed, scenario, and
year_effect. The latter is a data frame with columns year
and log_effect, giving the annual effect used by the simulator
before centring. It is metadata for checking the simulated example, not
an additional observed covariate. Centre it using the same reference
distribution as the diagnostic before comparing fitted year contrasts.