Modeling Rehabilitation of Mallard Ducks (Anas Plathyrhynchos) Using Logistic Regression

The use of statistics in a rehabilitation is a widely discussed topic among the wildlife community. In particular, the logistic regression can be an important tool to the wildlife community, if interpreted correctly. In this study, we used logit regression models to analyze trends in the rehabilitation of mallard ducks (Anas plathyrhynchos) in 2018 at a wildlife center based out of Southern California. We aimed to use the logistic regression as a tool for developing sampling models of wildlife rehabilitation regimes within the center. Seven models are reviewed using the intake characteristics: age, sex, and type of injury of the mallards (n = 1106) to predict the probability of deposition, released or euthanized. Two logit regression studies were conducted; one using euthanasia as the dependent variable (logiteuth), with the other one released as the dependent variable (logitrelease). Each study consisted of the same seven models. Model selection was based on the interpretation of the Akaike’s Information Criterion (AIC) as well as McFadden’s Pseudo R-Squared of the models. The most statistically significant model with an AIC of 279.19 in the logiteuth study utilized euthanasia as the dependent variable. From the logiteuth models, we were able to interpret the marginal effect and the probability of the model’s variables. The correlations between the reason for entry variables and the deposition variables confirm our hypothesis that orphaned mallards have the highest probability of release, while the injured mallards, limb injuries in particular were found to have the highest correlation with humane euthanasia. A proven statistical program, such as logistic regression, integrated into normal rehabilitation regimes, can benefit wildlife centers in preserving both finances and resources.