Statistical aspects of the analysis of group size effects in confined animals☆
Publication Type: |
Journal Article |
Year of Publication: |
2007 |
Authors: |
Mary C. Christman, Erin Hoerl Leone |
Publication/Journal: |
Applied Animal Behaviour Science |
Keywords: |
behavioral-responses, correlation, density effects, experimental design, group-size |
ISBN: |
01681591 |
Abstract:
In research on groups of animals housed in an enclosed environment, confounding is inevitably introduced when the factors under study include stocking density and group size. This confounding occurs because of the requirement that the space be limited and hence pen size and its associated physical attributes are altered in tandem with the levels of the factors. For example, when changing group size with one pen size, stocking density is altered simultaneous with group size. Similarly, when maintaining constant stocking densities at different group sizes, pen size is altered. Unfortunately this confounding of main effects is not always addressed or even acknowledged. This paper discusses the pitfalls associated with ignoring the collinearity among the effects and provides some guidance as to possible remedies. By examining several practical examples involving group size, stocking density, and pen attributes such as total area, perimeter space, and distance to walls and corners, we uncover the underlying issues related to the inability to fully distinguish between all effects of interest. Through the use of simulated data we address the analysis and interpretation of data containing confounded variables. While the factors stocking density, group size, and pen size/shape cannot all be made orthogonal concurrently, we offer several suggestions for minimizing the confounding that are practical and relevant to the study of behavior in groups of confined animals.