Skip to main content

Table 4 Results of simulations for data generated from a mixed model as in Fig. 3 with uncorrelated random effects and an unobserved confounder between \(M_1\) and \(Y_1\)

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

n

True

Estimates

Rel. bias

RMSE

Coverage rate

SDE

SIE

SDE

SIE

SDE

SIE

SDE

SIE

SDE

SIE

100

0.700

0.100

0.704

0.119

0.006

0.185

0.317

0.086

0.898

0.886

0.956

0.257

0.907

0.250

-0.051

-0.029

0.410

0.173

0.888

0.862

1.064

0.384

0.966

0.322

-0.092

-0.163

0.446

0.239

0.878

0.832

1.123

0.527

0.984

0.355

-0.124

-0.326

0.464

0.311

0.862

0.732

1.162

0.660

0.990

0.370

-0.149

-0.440

0.478

0.398

0.844

0.604

1000

-

-

0.697

0.122

-0.004

0.221

0.106

0.035

0.856

0.754

-

-

0.899

0.255

-0.059

-0.006

0.148

0.053

0.816

0.860

-

-

0.957

0.327

-0.101

-0.149

0.181

0.090

0.750

0.716

-

-

0.974

0.359

-0.132

-0.318

0.211

0.185

0.674

0.282

-

-

0.979

0.373

-0.158

-0.435

0.237

0.299

0.618

0.056

  1. For every model, each row refers to a different time \(t=1,\dots ,5\)