Enter the matrix: how to analyze the structure of behavior

Several methods are available for analyzing different aspects of behavioral transition matrices, but a comprehensive framework for their use is lacking. We analyzed parasitoid foraging behavior in environments with different plant species compositions. The resulting complex data sets were analyzed using the following stepwise procedure. We detected abrupt changes in the event log files of parasitoids, using a maximum likelihood method. This served as a criterion for splitting the event log files into two parts. For both parts, Mantel's test was used to detect differences between first-order transition matrices, whereas an iterative proportional fitting method was used to find behavioral flows that deviated from random transitions. In addition, hidden repetitive sequences were detected in the transition matrices on the basis of their relative timing, using Theme. We discuss the results for the example from a biological context and the comprehensive use of the different methods. We stress the importance of such a combined stepwise analysis for detecting differences in some parts of event log files.