Table 4.

Comparison of the estimated causal odds ratio using the trend-in-trend design and the cohort study method. The population is stratified into five and ten subgroups for the trend-in-trend algorithm. Confounders are sampled with autocorrelation 0.5 between any two consecutive calendar period. SD indicates standard deviation.

	Number of Strata	Number of Unmeasured Confounders	Trend-in-Trend Odds Ratio		Cohort Study Odds Ratio
			Mean (% bias)	SD*	Mean (% bias)	SD*
2.5	5	0	2.46 (−1.6)	0.0195	2.50 (0.0)	0.0097
		2	2.45 (−2.0)	0.0202	4.78 (89.2)	0.0098
		4	2.42 (−3.2)	0.0207	4.87 (94.8)	0.0098
	10	0	2.47 (−1.2)	0.0162	2.50 (0.0)	0.0097
		2	2.45 (−2.0)	0.0185	4.78 (89.2)	0.0098
		4	2.43 (−2.8)	0.0188	4.87 (94.8)	0.0098
2.0	5	0	1.98 (−1.0)	0.0176	2.00 (0.0)	0.0087
		2	1.94 (−3.0)	0.0185	4.23 (112)	0.0086
		4	1.94 (−3.0)	0.0172	4.30 (115)	0.0085
	10	0	1.97 (−1.5)	0.0141	2.00 (0.0)	0.0087
		2	1.93 (−3.5)	0.0150	4.23 (112)	0.0086
		4	1.94 (−3.0)	0.0144	4.30 (115)	0.0085
1.5	5	0	1.53 (2.0)	0.0119	1.51 (0.7)	0.0087
		2	1.52 (1.3)	0.0125	3.25 (117)	0.0086
		4	1.47 (−2.0)	0.0122	3.35 (123)	0.0091
	10	0	1.54 (2.7)	0.0105	1.51 (0.7)	0.0087
		2	1.53 (2.0)	0.0112	3.25 (117)	0.0086
		4	1.47 (−2.0)	0.0110	3.35 (123)	0.0091
1.0	5	0	1.02 (2.0)	0.0094	0.99 (−1.0)	0.0081
		2	1.02 (2.0)	0.0105	2.09 (109)	0.0079
		4	1.03 (3.0)	0.0107	2.20 (120)	0.0080
	10	0	1.03 (3.0)	0.0091	0.99 (−1.0)	0.0081
		2	1.04 (4.0)	0.0089	2.09 (109)	0.0079
		4	1.04 (4.0)	0.0092	2.20 (120)	0.0080