Abstract
Three problems which arise in the analysis of data from natural classroom settings are examined and the effects of the use of different statistical procedures are illustrated with data derived from the Home Environment and School Study in Canberra in 1969. These three problems involve: (1) the appropriate procedures to be used in the analysis of data when prior performance influences teacher and classroom behaviours and treatment conditions, (2) the appropriate units of analysis when data associated with schools, classrooms, and students within classrooms are analysed, and (3) the effects on statistical inference of using samples that involve clustered groups of students in classes or schools rather than simple random samples. These three problems are commonly ignored and can give rise to gross errors in the interpretation of the findings of school and classroom research studies.