META-ANALYSIS AND INTERVENTION LITERATURE: ISSUES OF DATA STRUCTURES (RESEARCH INTEGRATING, REVIEW, STATISTICS, DATA)
Abstract
The purpose of this study was to explore the data structures of four meta-analyses of intervention literatures in four distinct fields such that conclusions could be drawn regarding the impact of four aspects of data structures on meta-analytical strategies. The four targeted aspects of data structure were: (a) specificity of the unit of analysis, (b) normality of effect size distribution, (c) missing data for categorical independent variables, and (d) completeness of contingency tables for categorical independent variables. The finding of variability of values within study for a high frequency of categorical independent variables represents an obstacle to the use of an aggregated effect size for each independent study. The finding of leptokurtic and skewed effect size distributions when multiple effects were derived from single studies is problematic for the application of traditional inferential techniques to meta-analytical data sets. The finding of frequently missing data for categorical variables occurring in non-random patterns threatens the validity of meta-analytical findings and is troublesome for the handling of such missing data. The identification of patterns of confounded/dependent contingency relationships for categorical independent variables is problematic for demonstrating independent univariate effects. The effects of these and other structural aspects on traditional and homogeneity techniques of integration were explored. Based upon the findings of this study it is recommended that integrators of intervention research consider meta-analysis a form of complex survey research. The structural aspects of meta-analytical data sets should be explored for their substantive merits as well as for determining the impact they may have on analysis and interpretation. Directions for future research are presented.
Degree
Ph.D.
Subject Area
Psychology
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.