Skip to main content

Table 2 Estimation of the true effect in the real-world datasets using three different methods under the MCAR mechanism

From: Propensity score analysis with missing data using a multi-task neural network

Missing rate

Method

Mean

SD

RMSE

0.2

Missing indicator

352.262

250.792

431.108

Multiple imputation

198.329

276.932

576.882

Multi-task neural network

672.620

121.163

121.075

0.3

Missing indicator

395.330

283.958

415.239

Multiple imputation

403.878

309.952

425.198

Multi-task neural network

718.292

144.233

136.098

0.4

Missing indicator

316.312

226.881

450.458

Multiple imputation

277.395

247.175

493.797

Multi-task neural network

736.417

146.882

140.491

0.5

Missing indicator

240.517

266.086

534.726

Multiple imputation

341.053

279.437

455.591

Multi-task neural network

683.922

112.963

110.333

0.6

Missing indicator

339.664

233.459

433.169

Multiple imputation

171.271

191.980

570.923

Multi-task neural network

623.39

140.996

160.172

0.7

Missing indicator

323.533

202.261

433.415

Multiple imputation

318.219

232.415

451.292

Multi-task neural network

587.779

116.185

166.178

0.8

Missing indicator

328.838

128.502

402.568

Multiple imputation

395.226

123.355

338.146

Multi-task neural network

640.485

167.939

174.043

  1. SD, standard deviation; RMSE, root mean square error; MCAR, missing completely at random; MAR, missing at random; MNAR, missing not at random