A New Method for Analyzing Sequential Processes

Abstract

Researchers studying sequential processes (e.g., marital conflicts, teacher-student interactions, etc.) often try to model how recent events affect current events. A researcher doing so faces several difficulties: the threat of combinatorial explosion due to comprehensive coding, continuous and discrete variables, and differences across time (nonstationarity) and across groups (group heterogeneity). The authors discuss three often-used methods of analyzing time-series data (conditional probabilities, sequential analysis, and Logit with lag variables) and the problems inherent in them. The authors then introduce a new method that addresses the above problems: dynamic multilevel analysis. To highlight the similarities and differences between these methods, the authors apply them to data from student group problem-solving sessions in an algebra class. The authors use the various methods to show how likelihood of agreement was affected by other recent speakers’ correct ideas, mathematics status, agreement, and rudeness.