Results of model selection for negative binomial regression models of high or moderate severity (Class 1–2) bear conflicts, New York, USA, July–September 2006–2019.

We considered five candidate models: (a) no covariates (NULL), (b) mild severity (Class 3) conflicts early in the year (April–June) and bear abundance index (PRIOR+POP), (c) mild severity conflicts, abundance, and precipitation difference from means (PRIOR+POP+PRECIP), (d) mild severity conflicts, abundance, and parks (PRIOR+POP+PARK), and (e) a global model (PRIOR+POP+PRECIP+PARK). Models were ranked using Akaike Information Criterion for small samples (AICc).