Abstract
Resource selection functions (RSF) are flexible tools to estimate values proportional to the probability of use of a resource unit. The movement of virtual animals was simulated in a quadrat of 20 × 20 cells, each containing one of two resource categories for which the animals had certain preferences. Three different types of data were generated from the simulated tracks, and appropriate RSF models were applied to the data. Point estimates of the relative probability of selection were calculated to compare the performances of logistic, Poisson and Cox proportional hazard RSF models under a trade-off between tracking duration and sampling interval. A logistic model showed a systematic fall and rise of estimates for preferred and less preferred resources, respectively, due to an increase in the repeated use of the same cells. Longer sampling intervals alleviated the systematic bias but increased variation in the estimates. The Poisson model using count data demonstrated a constant difference between the estimates and the preference values that arose from the simulation settings of the selection process. The Cox proportional hazard model for survival time showed biases related to sampling interval. The use of simulation data is helpful for understanding the fundamentals of RSF models as well as for comparing the performances of different models.
