Table 1. Variation in the number of large carnivore attacks on humans over time and among species.
| COMPETING MODELS | β | SE | p | AICc | ΔAICc | Weighted AICc | |
|---|---|---|---|---|---|---|---|
| 1a | |||||||
| Year + Species + Year: Species | 850.5 | 0.52 | |||||
| Intercept | −0.333 | 0.453 | 0.462 | ||||
| Year | 0.024 | 0.010 | 0.016 | ||||
| Grizzly | 0.737 | 0.527 | 0.162 | ||||
| Black bear | 0.472 | 0.539 | 0.381 | ||||
| Cougar | 0.499 | 0.517 | 0.335 | ||||
| Wolf | 0.514 | 0.639 | 0.421 | ||||
| Coyote | 0.115 | 0.599 | 0.848 | ||||
| Polar bear | −0.398 | 1.093 | 0.716 | ||||
| Year: Grizzly | −0.009 | 0.012 | 0.458 | ||||
| Year: Black bear | −0.008 | 0.012 | 0.530 | ||||
| Year: Cougar | 0.006 | 0.011 | 0.609 | ||||
| Year: Wolf | −0.016 | 0.016 | 0.326 | ||||
| Year: Coyote | 0.023 | 0.013 | 0.071 | ||||
| Year: Polar bear | 0.002 | 0.023 | 0.930 | ||||
| Year + Species | 850.6 | 0.2 | 0.48 | ||||
| Species | 908.1 | 57.6 | 0.00 | ||||
| Year | 933.5 | 83.0 | 0.00 | ||||
| Null model | 996.3 | 145.8 | 0.00 | ||||
| 1b | |||||||
| Year + Species + Year: Species | 508.6 | 0.88 | |||||
| Intercept | 0.411 | 0.284 | 0.148 | ||||
| Year | 0.015 | 0.007 | 0.035 | ||||
| Cougar | −0.264 | 0.391 | 0.500 | ||||
| Coyote | −0.706 | 0.524 | 0.178 | ||||
| Year: Cougar | 0.015 | 0.009 | 0.104 | ||||
| Year: Coyote | 0.034 | 0.012 | 0.004 | ||||
| Year + Species | 512.6 | 4.0 | 0.12 | ||||
| Year | 529.4 | 20.8 | 0.00 | ||||
| Species | 549.7 | 41.1 | 0.00 | ||||
| Null model | 575.3 | 66.7 | 0.00 | ||||
(1a) Comparison of the five competing models built to study variation in the number of large carnivore attacks on humans over time and among species (n = 231). Summary of fitted parameters is shown for the most parsimonious candidate model (the selected model was the one with the lowest AICc score). Competitive models are ranked from the lowest AICc value (best model) to the highest one. (1b) We present the same analysis, but removed those species showing some patterns in the residuals of 1a. It is worth mentioning that in both cases we obtained the same results. European brown bear is included in the intercept. Negative binomial distribution error was selected over a Poisson distribution error, considering the output of the function odTest from the “pscl” package in R, which compares the log-likelihood ratios of a Negative Binomial regression to the restriction of a Poisson regression (critical value of test statistic at the alpha = 0.05 level: 2.7055; Chi–Square Test Statistic = 10.661, P < 0.001)