Survival regression analysis: a powerful tool for evaluating fighting and assessment

Theoretical models of animal contests frequently generate predictions about how asymmetries (e.g. differences in size, residence status) between conte stants affect fight duration. Linear regression and nonparametric correlati on analyses are commonly used to test the fit of data to such models. We sh ow how survival regression analysis (SRA) is a powerful technique for study ing the effect of asymmetries on the duration of contests. SRA, which is un der-utilized by students of animal behaviour, offers several advantages ove r more frequently used procedures. It provides unbiased parameter estimates even when including censored data (i.e. results of contests that have not ended at the time when observations are stopped). The analysis of hazard fu nctions, which is a component of SRA, is an easy way to test for consistenc y with predictions of the sequential assessment game model. These and other advantages of SRA are illustrated by using SRA and more conventional metho ds to analyse the effect of asymmetries on contest duration for encounters between female Mediterranean tarantulas, Lycosa tarentula (L.). It is hoped that this example of the advantages of SRA will encourage more widespread use of this powerful technique. (C) 2000 The Association for the Study of A nimal Behaviour.