Parametric Assumptions in Parasitoid Functional Response Analysis

ABSTRACT

The functional response of parasitoids is a key metric for evaluating their effectiveness as biological control agents and has been characterized for numerous host–parasitoid systems under diverse conditions. This study examines how parametric assumptions affect statistical inference in parasitoid functional response analysis, with particular attention to parameter estimation and model selection. Conventional approaches commonly use least squares or maximum likelihood methods based on heuristic distributions, such as the binomial or beta‐binomial distribution. From the perspective of point estimation, least squares can be viewed as equivalent to maximum likelihood under normally distributed errors with constant variance. Using simulations and published datasets, this study shows that arbitrary choices among these methods can lead to different estimates of functional response parameters and different conclusions about whether the response is type II or type III. In addition, a more flexible likelihood‐based approach is developed by characterizing the probability distribution of the data through simulations of the parasitism process. Comparisons with beta‐binomial maximum likelihood show that the simulation‐based model provides a better fit for some datasets, indicating that conventional heuristic distributions may not be flexible enough to describe the variability present in functional response data. The analyses also suggest that some datasets may contain insufficient information to distinguish reliably between competing functional response models, even when conventional model‐selection procedures appear to favour one model. These results show that variability in functional response data should be explicitly considered, because it can strongly influence parameter estimation and model selection.