WHY CLUSTERING MATTERS IN PEDAGOGICAL EXPERIMENTS: LESSONS FROM A PAIRED CLUSTER-RANDOMISED TRIAL WITH PRE-SERVICE TEACHERS

Pedagogical experiments in teacher education usually assign intact academic groups but analyse students as if they were independent. Using data from a paired cluster-randomised trial with 16 groups of pre-service primary teachers, this paper shows how the design effect, the intraclass correlation and the minimum detectable effect size change the precision of conclusions, and offers practical recommendations for dissertation research.