Methods for Synthesizing Regression Results

Florida
Associative
Synthesis
Research Methods
Principal Investigator: 
Betsy Jane Becker
Co-Investigator: 
Ingram Olkin
Project Overview
Background & Purpose: 

Our goal is to develop and study methods for summarizing the results of regression studies. This involves development of study indices, examination and study of existing indices, and considering the issues that arise when the regression models to be summarized differ across studies.

Research Design: 

We are developing and studying methods to analyze meta-analysis data. Our data are from simulation studies that inform about the behavior of statistical indices. We are examining ways of summarizing evidence from correlation and regression analysis via meta-analysis.

Findings: 

We have developed an index to represent the predictive value of a variable in a regression model and its variance. We have shown that a particular index of correlation among predictors is a maximum likelihood estimator. We have examined regression slopes that represent mean differences. We have examined dependence among effect sizes that share control groups, which relate to slopes for dummy variables.

Publications & Presentations: 

Becker, B. J., & Olkin, I. (under review). Combining correlation matrices from multiple studies. Manuscript under review at Journal of Educational and Behavioral Statistics.

Aloe, A. M., & Becker, B. J. (in press). Advances in combining regression results in meta-analysis. In W. P. Vogt & M. Williams (Eds.), Handbook of methodological innovation. London: Sage.

Becker, B. J. (2009). Model based meta-analysis. In H. M. Cooper, L.V. Hedges & J. Valentine (Eds.), The handbook of research synthesis. (2nd ed.). New York: Russell Sage Foundation.

Becker, B. J., & Wu, M.J. (2008). Generalizability and research synthesis. In K. Ercikan and W-M. Roth (Eds.), Generalizing in education research. Lawrence Erlbaum.

Kim, R., & Becker, B. J. (under revision). The degree of dependence between multiple-treatment effect sizes. Manuscript under revision for Multivariate Behavioral Research.

Aloe, A. M., & Becker, B. J. (2007, November). Meta-analysis and effect sizes from t statistics. Paper presented at the annual meeting of the Florida Educational Research Association, Tampa, FL

Becker, B. J., & Olkin, I. (2008, March). A maximum likelihood estimator of the pooled correlation matrix. Paper presented at the annual meeting of the American Educational Research Association, New York, NY.

On the Web: 
http://mailer.fsu.edu/~bbecker/projects.htm