What Is Wrong With ANOVA and Multiple Regression? Analyzing Sentence Reading Times With Hierarchical Linear Models

• Published in: Discourse processes Vol. 41; no. 3; pp. 221 - 250
• Main Author: Richter, Tobias
• Format: Journal Article
• Language: English
• Published: Philadelphia Lawrence Erlbaum Associates, Inc 01.01.2006; Taylor & Francis Ltd
• Subjects: Causal Models; Computer Software; Data analysis; Guidelines; Hypothesis Testing; Predictor Variables; Reading; Reading Rate; Reading Tests; Regression (Statistics); Regression analysis; Research Design; Research Subjects; Sentence structure; Sentences; Statistical Analysis; Time; Variance analysis
• ISSN: 0163-853X; 1532-6950
• DOI: 10.1207/s15326950dp4103_1

Most reading time studies using naturalistic texts yield data sets characterized by a multilevel structure: Sentences (sentence level) are nested within persons (person level). In contrast to analysis of variance and multiple regression techniques, hierarchical linear models take the multilevel structure of reading time data into account. They provide methods to estimate variance components and to model the influence of predictor variables on different levels as well as cross-level interactions between these predictors. This article gives a brief introduction to the method and proposes practical guidelines for its application to reading time data, including a discussion of power issues and the scaling of predictor variables. The basic principles of model building and hypothesis testing are illustrated with original data from a reading time study with naturalistic texts.