Multivariate data handling in the study of rat behavior: an integrated approach

The aim of the present article is to provide a methodological description of various approaches to multivariate data handling in the study of rodent behavior. To this purpose, 42 male Wistar rats were tested in an open field, and their behavior was recorded through a digital video camera for a subsequent analysis by means of a software coder. After a preliminary evaluation of descriptive features such as durations and percent distributions, we carried out different kinds of multivariate approaches represented by stochastic, cluster, adjusted residual, and T-pattern analyses. In the attempt to depict behavior in a straightforward way, the results of each analysis were graphically illustrated through path diagrams, dendrograms, histograms, and tree-shaped T-patterns. Path diagrams showed a clear behavioral convergence toward immobile sniffing; dendrograms highlighted three different dyadic clusters: walking/climbing, immobile-sniffing/immobility, and paw-licking/grooming; adjusted residuals confirmed, for the same patterns, highly significant association values; finally, T-pattern analysis showed a highly recurring temporal sequence of events encompassing walking, climbing, immobile sniffing, and immobility. Such results, drawing attention to specific behavioral patternings, strengthen and extend previous findings on rodent behavior. We suggest that T-pattern analysis, integrated with other multivariate approaches, can provide a more detailed and complete rat behavior representation, very different from classical quantitative