18.4 Other randomizations

• Matched pairs: Special case of randomized blocks where there are just 2 in each block, randomly one of the pair receives the treatment.

• Group / Cluster randomized experiments

  • Assign treatments at group level.
  • Sometimes a design decision to avoid interference effects (Violation of SUTVA)
  • Can analyze at group level (aggregated measures) or use multilevel regression (Book 2, in progress? “Applied Regression and Multilevel Models”)