A Nested Frailty Survival Approach for Analyzing Small Group Behavioral Observation Data

Abstract

Patterns of behavior during social interaction have long been of interest to small group researchers. Within social interaction, the probability of an initial behavior meeting with a particular response (e.g., an angry comment meeting with an angry rejoinder) often depends, in part, on the duration of the initial behavior. This article presents a nested frailty approach that accounts for the duration of individual behavior and is well suited to the examination of small group data. While traditional sequential techniques disregard information about the duration of behaviors, survival methods are capable of modeling behavioral duration in sophisticated ways. Expanding on previously proposed survival methods for behavioral observation research, the nested frailty approach involves three levels of hierarchical clustering and is thus well suited to analyzing data from a variety of different social configurations. An example application explores the spreading of smiles within 160 groups of three in a laboratory-based social interaction.