Characterizing the Sequential Structure of Interactive Behaviors Through Statistical and Grammatical Techniques

Abstract

Statistical and grammatical techniques are reviewed as an integrated approach to exploratory sequential data analysis (ESDA) for categorical data. The first step is the identification and validation of the categories to be analyzed. The main statistical techniques discussed are log-linear modeling and lag sequential analysis. These methods allow for the statistical evaluation of a wide range of general and specific hypotheses about sequential structure. Grammatical techniques based on definite-clause grammars are described and illustrated, and the complex issue of measuring the goodness of fit of a set of patterns is discussed. Throughout the article, examples from our own research illustrate how the various techniques are used, especially in concert, while carrying out ESDA. In Section 6, several other human-computer interaction and computer-supported cooperative work applications of these techniques are discussed.