Re-expressing coefficients from regression models for inclusion in a meta-analysis

Table 3 Performance measures based on simulation scenarios with stated values of σ, median, and log base for a total of 8 simulation scenarios^a (n_sim = 2000 per scenario^b)

σ	Median	Log base	Estimator	Average relative bias	Average coverage probability (%)^c
0.45	2	2	β_RB	-0.008	99.1
			β_Dz	0.199	78.7
			β_Alt	0.2	78.6
0.45	2	10	β_RB	0.021	99.9
			β_Dz	0.231	99.9
			β_Alt	0.019	99.9
0.45	16	2	β_RB	0.182	81.9
			β_Dz	0.195	79.5
			β_Alt	0.195	79
0.45	16	10	β_RB	0.21	100
			β_Dz	0.224	99.9
			β_Alt	0.012	100
0.85	2	2	β_RB	0.655	2.4
			β_Dz	0.901	0.2
			β_Alt	1.003	0
0.85	2	10	β_RB	0.678	56.7
			β_Dz	0.927	37.4
			β_Alt	0.676	57.1
0.85	16	2	β_RB	0.979	0
			β_Dz	0.898	0.1
			β_Alt	1	0
0.85	16	10	β_RB	0.987	31.4
			β_Dz	0.91	36.9
			β_Alt	0.658	56.9

^aFor all scenarios used in this table the n_obs was 162 and the β_DGM was 1
^bThe median Monte Carlo standard error of the relative bias was ≤ 0.003 for all estimation methods
^cCalculated as the sum of the scenarios where the re-expression method confidence interval included the observed β (β_estimand) divided by the total number of scenarios (960)

ISSN: 1471-2288