Table 3.

Simulation Results: IE estimates for three datasets.

		Dataset 1			Dataset 2			Dataset 3
		Truth	IE	Bias	Truth	IE	Bias	Truth	IE	Bias
Age	1	−1	−0.997	0.003	−1	5.747	6.747	−3	0.249	3.249
	2	0	−0.002	−0.002	0	0.002	0.002	0	0.000	0.000
	3	1	0.999	−0.001	1	−5.749	−6.749	3	−0.249	−3.249
Period	1	−7	−6.999	0.001	−7	−13.75	−6.750	−1	−4.250	−3.250
	2	0	−0.002	−0.002	0	0.002	0.002	0	−0.002	−0.002
	3	7	7.002	0.002	7	13.748	6.748	1	4.252	3.252
Cohort	1	−2	−2.001	−0.001	−20	−6.497	13.503	−8	−1.500	6.500
	2	−1	−0.998	0.002	−10	−3.253	6.747	−4	−0.750	3.250
	3	0	−0.001	−0.001	0	0.002	0.002	0	0.000	0.000
	4	1	1.004	0.004	10	3.250	−6.750	4	0.750	−3.250
	5	2	1.996	−0.004	20	6.498	−13.502	8	1.500	−6.500

Notes:

• 1
  For each dataset, the IE estimates are averaged over 1,000 simulations.
• 2
  The bias of IE is evaluated by the difference between the true effects and the IE estimates, averaged over 1,000 simulations.
• 2
  Equation (17) holds for dataset 1, but does not hold for datasets 2 and 3.