Skip to main content

Table 3 Bootstrap estimates for each imputed dataset (combination estimate in Table 2 is mean value across imputations, standard error follows from Rubins’s rules

From: G-estimation of causal pathways in vocational rehabilitation for adults with psychotic disorders – a secondary analysis of a randomized trial

 

COGNt

log (PANSSsumt)

log (PANSSnegt)

 

\( \kern0.75em {\hat{\psi}}_0^{imp,b} \)

\( st. error \)

\( {\hat{\psi}}_0^{imp,b} \)

\( st. error \)

\( {\hat{\psi}}_1^{imp,b} \)

\( st. error \)

\( {\hat{\psi}}_0^{imp,b} \)

\( st. error \)

\( {\hat{\psi}}_1^{imp,b} \)

\( st. error \)

Imputation

1

0.216

0.08

1.839

2.752

−0.294

0.11

3.686

1.901

−0.3

0.086

2

0.207

0.076

−0.151

2.539

−0.194

0.099

3.084

1.791

−0.249

0.076

3

0.212

0.078

2.128

2.687

−0.331

0.11

3.805

1.965

−0.254

0.089

4

0.234

0.078

1.275

2.668

−0.335

0.126

3.704

1.936

−0.317

0.089

5

0.230

0.08

0.057

2.643

−0.245

0.109

2.632

1.897

−0.248

0.081

6

0.245

0.077

2.436

2.894

−0.344

0.128

4.52

2.025

−0.347

0.095

7

0.252

0.077

0.925

2.636

−0.332

0.116

3.055

1.85

−0.266

0.081

8

0.196

0.078

2.233

2.749

−0.326

0.122

4.334

2.078

−0.346

0.097

9

0.237

0.077

1.348

2.778

−0.304

0.122

3.348

1.881

−0.253

0.082

10

0.212

0.076

3.412

2.816

−0.39

0.129

3.735

1.978

−0.297

0.085

ra

0.058

0.178

0.242

0.099

0.226

FMIb

0.055

0.151

0.195

0.09

0.184

  1. a: relative increase in variance due to missingness, b: fraction of missing information