Lag sequential analysis: Robust statistical methods.

Stephen V. Faraone,Donald D. Dorfman

DOI: https://doi.org/10.1037/0033-2909.101.2.312

IF: 22.4

1987-01-01

Psychological Bulletin

Abstract:Lag sequential analysis (Sackett, 1979, 1980) has become an important tool for researchers of interpersonal interaction. This method enables one to explore and summarize cross-dependencies occurring in complex interactive sequences of behavior. Statistical methods for lag sequential analysis have been found to be incorrect (Allison & Liker, 1982) or make strong theoretical assumptions difficult to verify. In this article we clarify confusion about statistics proposed by Gottman (1979a, 1979b) and Sackett (1979). We then present a statistical test of cross-dependency that derives from the assumption that each behavioral sequence is a first-order Markov chain. Lastly, we introduce two robust methods—the jackknife and the data split—for testing cross-dependence and estimating confidence intervals about indices of cross-dependence. We use Monte Carlo simulations to demonstrate the adequacy of these methods and to provide a basis for choosing between them.

psychology, multidisciplinary