Estimating the effect of class size on academic achievement by ex post facto experiment

The analyses of influence of class size on academic achievement used data from study conducted in 2006 by the Regional Examination Board in Cracow (Poland). The variables explaining the achievements of lower secondary school pupils were identified using regression analysis. The model explains 71% of variance of exam results. These variables were used to identify statistical twins. Their assignment to the experimental and control group was performed in three ways: by stratification using Mahalanobis distance, matching one-to-many and one-to-one using k-means method. The last method proved the most successful. The effect of class size on student outcomes proved statistically insignificant. However, pupils from classes with below 23 pupils achieved higher mean scores than their peers from larger classes by 0.039 standard deviation.

http://www.edukacja.ibe.edu.pl/images/numery/2013/6-3-koniewski-estimating-the-effect-of-class-size.pdf