G.P. Sackett and J.C. Ha
Regional Primate Research Center, University of Washington, Seattle, USA
Multiple spell, discrete-time, survival analysis (MDSA) can be used to statistically study behaviors that rarely occur in a temporal stream of observational data. Examples include bouts of aggression among animal or child subjects, seizures among developmentally disabled individuals, attainment of milestones in a developmental processes such as object permanence or a motor reflex, or the onset of bouts of sexual activity in a group of primates. Using standard parametric or nonparametric techniques with such data to study group differences or the effects of covariates can be problematic because (1) particular individuals may not be available for observation on all sessions a problem called censoring in survival analysis terminology, (2) session lengths may differ both within and between individuals, (3) relatively small and unequal sample sizes may invalidate parametric assumptions, and (4) some individuals may never exhibit the behavior in question. MDSA can be used to study the effects of particular independent variables and covariates under any or all of these conditions.
This presentation will describe the statistical methods used in MDSA. Am example analysis concerning aggression as a function of sex, dominance rank, troop size, and prior year rainfall will be used to illustrate the application of MDSA.