TEACHING STYLES AND PUPIL PROGRESS: A RE‐ANALYSIS

Abstract

S ummary . The object of the study was to examine the statistical techniques available for the analysis of process‐product studies involving non‐randomised quasi‐experimental designs, and to demonstrate the practical effects of their use on the data from the Teaching Styles study (Bennett, 1976). Of particular concern were the ‘unit of analysis’ or aggregation problem, and the differential effects of treatment grouping by cluster and factor methods. The original grouping of teachers into formal, informal and mixed styles was investigated using a latent class model for the 38 binary questionnaire items. Convincing evidence of three overlapping latent classes was found. The comparison of latent classes in terms of pre‐test gain scores was examined using a series of variance component models, allowing for correlation of children within the same class. Differences among classes were altered by the probabilistic clustering of the latent class model compared to the original findings, and the significance of the differences was reduced when the correlation among children was allowed for.