Model selection results for comparison of random effects in models for absolute travel speeds.

Comparison criteria include number of parameters (k), negative log-likelihood (−2·ln(L)), and the Bayesian Information Criterion (BIC).
aIn all models, fixed parameters included additive effects of species, wild or hatchery-rearing history, Fraser River or non-Fraser origin (FnF), and fork length (FL). For the rivers dataset, a FnF:FL interaction was also included. There was also one fewer species group (Chinook) and one fewer rearing history group (‘unknown’) in the coastal dataset compared with the rivers dataset, hence three fewer fixed parameters overall in the coastal dataset. Models differ in their random effects considered: (γ0 k), random intercepts for years; (ω0 j), random intercepts for watersheds nested within Fraser/non-Fraser River; (ω0 j, γ0 k), random intercepts for both effects.