A Demonstration of the Three-Level Hierarchical Generalized Linear Model Applied to Educational Research

• Other Authors: Subedi, Bidya Raj (authoraut)
• Format: Others
• Language: English; English
• Published: Florida State University
• Subjects: Educational psychology
• Online Access: http://purl.flvc.org/fsu/fd/FSU_migr_etd-1521

This study demonstrates the three-level hierarchical generalized linear model (HGLM) applied to educational research. The sequential steps for developing, analyzing, evaluating, and applying the three-level HGLM are demonstrated in the study. In the study, the effects of predictors are interpreted using the simple effect and ANOVA-like approaches. In order to describe predictors' effects, odds and odds ratios are computed and interpreted. This study used NAEP 2000 Reading data for fourth grade students. A sample of 7,175 students, 1,076 teachers, and 295 schools from 46 States was used in the study. Student, teacher, and school level data were used as level-1, level-2, and level-3 units respectively for analysis. Reading proficiency was considered as a dichotomous outcome. Socioeconomic status (SES) and minority were used as student level predictors; class size was used as a teacher level predictor; and school mean SES was used as a school level predictor. Positive effect of SES and school mean SES on reading proficiency was found. However, negative effect was found due to minority and class type on reading proficiency. Graphical methods are presented to interpret the effects for class type and minority on reading proficiency. Specifically, the effect of class type is presented graphically for minority and non-minority students associated with different levels of school mean SES. Similarly, the effect of minority is depicted for crowded and non-crowded class types associated with different levels of school mean SES. The research practitioners not only can replicate the procedural steps of demonstrating the three-level HGLM as presented in this study, but they also can interpret predictors' effects using simple effect and ANOVA-like approaches described in this study. Despite the complexity of the process in computing effects using the simple effect approach, researchers can interpret effects with less complication using this approach compared to the traditional HGLM approach. === A Dissertation Submitted to the Department of Educational Psychology and Learning Systems in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy. === Spring Semester, 2005. === December 8, 2004. === Teacher and school effects, Interaction effect, Main effect, Odds and odds ratios, Unconditional and conditional models, NAEP reading data, Three-level HLM and HGLM, Cross-level interactions, Simple effect description === Includes bibliographical references. === Richard L. Tate, Professor Directing Dissertation; Janice Flake, Outside Committee Member; Albert Oosterhof, Committee Member; Akihito Kamata, Committee Member.