Methods of integrating observational and traditional data analysis

This essay argues that there is an insufficient correspondence between the epistemological assumptions which guide interaction analysis and the statistical assumptions which characterize the procedures most often used to analyze interaction data. The paper demonstrates that by combining Markov chain or lag sequential analysis with more traditional, parametric approaches (multiple regression, ANOVA, etc.), a much closer correspondence is achieved between the substantive questions posed by interaction analysts and the answers provided by existing approaches to data analysis.