- 01 Stratified Randomization Design Improves covariate balance between treatment and control groups by randomizing separately within pre-defined subgroups
- 02 Pairwise Matching & Re-randomization Maximises balance by pairing units on observed characteristics before treatment assignment or by rerunning randomization until balance criteria are met
- 03 Cluster Randomization Design Assigns treatment to groups rather than individuals when individual-level randomization is infeasible or contamination is a concern
- 04 Factorial & Multi-Arm Design Tests multiple treatments or treatment combinations simultaneously by assigning units to all possible combinations of factor levels
- 05 Stepped Wedge Design Rolls out treatment sequentially across clusters over time, with all clusters eventually treated and each serving as its own control before treatment begins
- 06 Encouragement / Promotion Design Estimates the causal effect of a program when take-up cannot be compelled — by randomly assigning encouragement to participate and using it as an instrument
- 07 Spillover & Saturation Design Estimates both direct treatment effects and spillovers onto untreated individuals by varying the density of treatment across clusters