Skip to main content

Table 5 The NHEFS data analysis: estimated average treatment effect of quitting smoking on weight gain (not quitting smoking as reference)

From: Estimation of average treatment effect based on a multi-index propensity score

Estimator

Estimates

BS-SE

95%-CI

P-value

Single model-based estimators

 IPW.model1

3.015

0.522

(1.992, 4.038)

 < 0.001

 IPW.model2

3.140

0.515

(2.131, 4.149)

 < 0.001

 IPW.ANN

2.404

0.560

(1.306, 3.502)

 < 0.001

 OR.model1

3.187

0.471

(2.264, 4.110)

 < 0.001

 OR.model2

3.254

0.477

(2.319, 4.189)

 < 0.001

 OR.ANN

3.392

0.825

(1.775, 5.009)

 < 0.001

Artificial neural network-based MiPS estimator

 MiPS-1000

2.713

0.510

(1.713, 3.713)

 < 0.001

 MiPS-0100

2.871

0.510

(1.871, 3.871)

 < 0.001

 MiPS-0010

2.584

0.468

(1.667, 3.501)

 < 0.001

 MiPS-0001

2.221

0.476

(1.288, 3.154)

 < 0.001

 MiPS-1100

2.880

0.505

(1.890, 3.870)

 < 0.001

 MiPS-1010

2.764

0.508

(1.768, 3.760)

 < 0.001

 MiPS-1001

2.704

0.520

(1.685, 3.723)

 < 0.001

 MiPS-0110

2.834

0.513

(1.829, 3.839)

 < 0.001

 MiPS-0101

2.868

0.520

(1.849, 3.887)

 < 0.001

 MiPS-0011

2.606

0.468

(1.689, 3.523)

 < 0.001

 MiPS-1110

2.847

0.515

(1.838, 3.856)

 < 0.001

 MiPS-1101

2.890

0.528

(1.855, 3.925)

 < 0.001

 MiPS-1011

2.868

0.546

(1.798, 3.938)

 < 0.001

 MiPS-0111

2.854

0.536

(1.803, 3.905)

 < 0.001

 MiPS-1111

2.873

0.526

(1.842, 3.904)

 < 0.001

  1. BS-SE bootstrapping standard error based on 500 resamples, 95%-CI 95% Wald confidence interval. The artificial neural network-based MiPS estimator which contains propensity score model and/or outcome regression model is denoted as “method-0000”, where each digit of the four numbers, from left to right, indicates if propensity score model 1, propensity score model 2, outcome regression model 1, outcome regression model 2 is included in the estimator (“1” indicates yes and “0” indicates no)