Modelling concerns confound evaluations of legal wolf-killing

Letter to the editor at https://www.elsevier.com/locate/biocon

Liberg et al. (2020)’s interesting article on the effect of legal killing on poaching and disappearance of wolves continued the seminal work of Liberg et al. (2012). However, our eyes were caught by unusual modeling procedures, so we reanalyzed their population model and found issues with the specification of their individual-level model. At the population-level, the authors modeled “Disappearance rate in relation to… number of wolf territories” (Fig. 2A) or in relation to “culling rate” (Fig. 2B) using logistic regressions (Table 1). We could not find probability estimates or logistic curves in those data depictions and overall had difficulty interpreting their methods. We think they analyzed annual population-wide disappearance rate (number of wolves that were poached or lost, divided by the number of territories variously referred to as population size or n of wolves at risk/2) as a continuous, not binary, variable. Using the population-level data from their appendix, we quantified the annual absolute number of wolves that disappeared for correlation with the absolute number culled (n = 17, r = 0.69) and as rates (divided by the annual number of wolves at risk and standardized as the authors did, r = 0.56), although the latter is unnecessary because the denominator was the same for both variables…..

(download document below to read the full letter)