TABLE 2.

Type I error and power (%) across simulation scenarios comparing an unadjusted analysis, adjusting for the randomisation strata and adjusting for the true strata when the treatment effect is of interest and all stratification errors are discovered.

			Type I error (%) (${\beta }_T=0$)			Power (%) (${\beta }_T=0.2$)
Stratification error rate	Errors across strata	Covariate effect (β X)	Unadjusted	Randomisation strata	True strata a	Unadjusted	Randomisation strata	True strata a
1%	Equal	1	2.85	4.90	4.85	83.05	88.13	88.40
1%	Equal	3	0.07	4.92	5.12	36.64	85.59	88.80
1%	Unequal	1	3.13	5.38	5.39	82.96	87.89	88.30
1%	Unequal	3	0.05	4.66	5.01	35.94	85.48	88.11
20%	Equal	1	4.21	5.09	4.98	81.80	83.79	88.37
20%	Equal	3	2.74	5.68	4.68	39.81	51.88	88.54
20%	Unequal	1	4.04	4.74	4.71	81.78	83.86	89.10
20%	Unequal	3	2.16	5.00	5.25	40.31	52.89	88.87

Note: The maximum Monte Carlo SE across all methods and scenarios was 0.23 for type I error and 0.50 for power.

^a Matches the updated strata in this setting where all stratification errors are discovered.