Hunt for Randomized Experimental Reports: Document Search and Efforts for a "What Works?" Meta-Analysis

• NCJ Number: 160571
• Journal: Journal of Crime and Justice Volume: 18 Issue: 2 Dated: (1995) Pages: 63-80
• Author(s): A J Petrosino
• Date Published: 1995
• Length: 18 pages
• Grant Number(s): 93-IJ-CX-0011
• Agencies: NIJ-Sponsored
• Publication Type: Issue Overview

Annotation

This paper discusses the process involved in searching, locating, and retrieving documents for a meta-analysis of crime- reduction field experiments.

Abstract

The inclusion criteria, the rationale for the searches used, the problems encountered, and a presentation of major techniques, including electronic technology, are discussed. Meta-analysis, once referred to by originator Gene Glass as the "analysis of analyses" or the quantitative analysis of research reports, has become a popular research tool in social science. The most frequent use of meta-analysis has been in the area of treatment effectiveness, where 13 studies that have examined the efficacy of crime-reduction programs have been reported. Meta-analysis has brought attention to the importance of systematizing the review process. All research reviews should be explicit, not only regarding findings, but also concerning how studies were chosen and how they were obtained. Perhaps most crucial to search efforts is the selection and refinement of inclusion criteria; this provides a set of guidelines and boundaries for data collection. This stage can be improved in a number of ways. Published solicitations and mail campaigns of prominent authors should be done early in the project. This would not only help in the retrieval process but would also enable the investigator to publicize the goals and objectives of the meta-analytic project to the research community. Prospective meta-analysts should also retrieve prior research reviews and published bibliographies, particularly those that are comprehensive in their coverage of the targeted subject area. Good literature reviews often generate hundreds of citations; therefore, good recordkeeping is essential in meta-analysis. As citations are checked, two computer files are essential: a potentially relevant documents file and a file for rejected documents. 48 references and an appended listing of dialog databases searched