Table 3.

Estimated log risk ratio for neonatal mortality due to RF and mean square error under varying true direct and indirect effects of RF on neonatal mortality, and varying direct effects of U on neonatal mortality; the effect of RF on birthweight is fixed at -200 g and the effect of U on birthweight is fixed at -600 g

Parameter values			True effects			Estimates - ADJUSTED MODELa			Estimates - UNADJUSTED MODELb
bc	dd	ee	ΘIf	ΘDg	ΘTh	β̂RF|BW	β̂RF|BW-θD	MSE	β̂RF	β̂RF-θT	MSE
	0		0	0	0	-0.054	-0.054	0.020	-0.001	-0.001	0.015
	0.2	0	0.2	0.2	0	0.148	-0.052	0.018	0.200	0.000	0.013
0.4	0.4		0.4	0.4	0	0.347	-0.053	0.016	0.400	0.000	0.011
	0		0.2	0	0.2	-0.069	-0.069	0.091	0.193	-0.007	0.072
	0.2	-0.0005	0.4	0.2	0.2	0.129	-0.071	0.079	0.388	-0.012	0.064
	0.4		0.6	0.4	0.2	0.334	-0.066	0.069	0.594	-0.006	0.054
	0		0	0	0	-0.314	-0.314	0.111	-0.004	-0.004	0.011
	0.2	0	0.2	0.2	0	-0.111	-0.311	0.108	0.199	-0.001	0.009
1.6	0.4		0.4	0.4	0	0.089	-0.311	0.107	0.399	-0.001	0.008
	0		0.2	0	0.2	-0.350	-0.350	0.183	0.190	-0.010	0.052
	0.2	-0.0005	0.4	0.2	0.2	-0.145	-0.345	0.171	0.391	-0.009	0.044
	0.4		0.6	0.4	0.2	0.060	-0.340	0.163	0.598	-0.002	0.039

^aAdjusted model includes birthweight (BW) and risk factor (RF) as independent variables.

^bUnadjusted model includes only RF as an independent variable.

^cb = log (p[NM = 1|U = 1]/p[NM = 1|U = 0]).

^dd = log (p[NM = 1|bw = k + 1]/p[NM = 1|bw = k]).

^ee = E[bw|RF = 1] - E[bw|RF = 0] = -200 g.

^fΘ_I: the indirect effect of RF=d * e.

^gc ≡ Θ_D=log (p[NM=1|RF=1]/p[NM=1|RF=0]).

^hΘ_T: the total effect of RF=c+(d * e).