14  Design methods ✎ Outline

14.1 Recruitment rates

• Plausibility of reported recruitment rates

14.2 Attrition rates

• Plausibility of reported attrition/dropout rates

14.3 Randomisation methods

14.3.1 Differences at baseline

• Carlisle method
• Baseline balance checks and variants

14.3.2 Randomisation plausibility

• E.g., claimed to randomise students to groups for an intervention, but it would have to be at the level of classrooms or schools
• Binomial/multinomial distribution checks: only useful for testing weirdly unbalanced assignment, not too-perfectly balanced, as it’s technically the most probable outcome