Longitudinal mediation analysis with multilevel and latent growth models: a separable effects causal approach

Table 5 Results of simulations for data generated from an LGM as in Fig. 4 with an unobserved confounder between \(\eta _{{0M}}\) and the outcome at all times. For every model, each row refers to a different time \(t=1,\dots ,5\)

n	True		Estimates		Rel. bias		RMSE		Coverage rate
n	SDE	SIE	SDE	SIE	SDE	SIE	SDE	SIE	SDE	SIE
100	0.140	0.000	0.124	0.000	-0.017	0.000	0.256	0.000	0.946	0.000
	0.410	0.160	0.399	0.163	-0.026	-0.019	0.340	0.178	0.958	0.966
	0.680	0.319	0.675	0.326	-0.007	-0.019	0.548	0.356	0.948	0.966
	0.950	0.479	0.951	0.488	0.001	-0.019	0.788	0.534	0.948	0.966
	1.220	0.639	1.227	0.651	0.006	-0.019	1.038	0.712	0.952	0.966
1000	-	-	0.140	0.000	0.002	0.000	0.079	0.000	0.938	0.000
	-	-	0.414	0.159	0.010	-0.003	0.098	0.041	0.944	0.964
	-	-	0.688	0.318	0.011	-0.003	0.152	0.082	0.948	0.964
	-	-	0.961	0.477	0.012	-0.003	0.217	0.122	0.960	0.964
	-	-	1.235	0.637	0.012	-0.003	0.284	0.163	0.958	0.964

ISSN: 1471-2288