Multimetric statistical analysis of behavior in mice selected for high and low levels of isolation-induced male aggression.

Behavioral observations as a matrix of probabilistic changes of postures and acts are multiple measurements that could introduce variability to statistical analysis. We propose the multimetric statistical algorithm that supplements the linear analysis of variance by pair correlation, factor and discriminant function analyses. Although these methods were utilized mostly in behavioral studies, the combined use in frame of one behavioral test was not done before. In present study statistical techniques were applied to analyze social behavior in Turku aggressive (TA) and Turku non-aggressive (TNA) mouse lines, bidirectional selected for offensive aggression towards an unknown male. Each statistical technique amplified new details of mouse behavioral profiles that give possibility to describe TA and TNA subjects in terms of Cloninger's model of personality. Also, it was identified that TA mice displayed fighting-biting aggression while TNA mice demonstrated immobile defensive strategy. Hypothetical discriminant formula was found for each mouse behavioral genotype that might be used to identify behavioral profile and line affiliation of unknown subjects.