Loading…

Can pre-experiment behavior improve experiment precision?

Compare raw and baseline-adjusted differences on the same assigned units. Declare the observation window or study design before running. The result includes denominators and leaves unidentified quantities uncomputed.

Required data: Assignment unit ID · Treatment (0/1) · Baseline outcome · Outcome

One pre-treatment covariate with ANCOVA2 and HC3. Post-outcome covariate or stopping choices invalidate interpretation.

Next step

Did the ads really work?

Common next stepPrepare a new dataset
Geo experiment & synthetic control
Find other analyses