Parameter estimates for best-supported Hurdle Negative Binomial (HNB) models predicting the probability of foraging behaviours of newly feeding Arctic charr (Salvelinus alpinus).

All fish were kept in isolation from hatching. Both outcomes of the models are reported: the first part (“Binomial error distribution”) is a binomial probability model that governs the binary outcome of whether a count variable has a zero (no foraging) or a positive value (foraging) and the second part (“Negative binomial error distribution”) is a zero-truncated count model that governs the positive outcomes (tested the importance of the independent variable only among the feeding fish). The variable “Total foraging” is the sum of all foraging attempts at all locations in the water column: bottom foraging, water column and surface foraging (see note).