A Comparison of Analysis of Covariance to Within-Class Regression in the Analysis of Non-Equivalent Groups

Publisher Website
Google Scholar
Abstract
Comparing non-equivalent groups is a persistent problem in educational research methodology, especially teacher effectiveness research. Within-class regression is a method, developed in this paper, of comparing a large number of non-equivalent groups. Monte Carlo data were generated under several conditions and within-class regression. The results indicated that the within-class regression method was a less biased method of data analysis. Reading achievement data were also analyzed using both methods. The results indicated that the method of analysis makes a difference in analyzing treatment effects. It was concluded that, when a large number of non-equivalent groups are compared, within-class regression will yield more accurate estimates of treatment effects than analysis of covariance.

This publication has 6 references indexed in Scilit: