Random and Fixed Effects Models in Meta-Analysis in Criminal Justice and Criminology: Choosing the Appropriate Method

• NCJ Number: 186499
• Journal: Journal of Crime and Justice Volume: 23 Issue: 1 Dated: 2000 Pages: 65-79
• Author(s): Travis C. Pratt
• Editor(s): J. Mitchell Miller
• Date Published: 2000
• Length: 15 pages
• Publication Format: Article
• Publication Type: Research (Applied/Empirical)
• Language: English
• Country: United States of America

Annotation

In recent years, meta-analytic reviews of empirical literature have become increasingly common in criminal justice and criminology; most studies, however, contain multiple statistical models and therefore produce multiple effect size estimates of the same relationships.

Abstract

The lack of statistical independence in such cases may produce estimation errors in a meta-analysis since effect size estimates within each study may be similar due to their production from a common source. Random and fixed effects models have been developed to correct for this potential lack of statistical independence across effect size estimates. While both random and fixed effect methods are useful in certain settings, determining which method is appropriate in a given situation for a meta-analysis is less clear. The author presents a framework for determining which method is appropriate to correct for independence problems when subjecting different types of research in criminal justice and criminology to a meta-analysis. 34 references and 7 notes