Table 3 Comparison of the distributed model with the centralised model and with the local model (mean for all model and all columns).

From: Evaluating distributed-learning on real-world obstetrics data: comparing distributed, centralized and local models

  

M

SD

95% CI

P

AUPRC

distributed

0.691

0.216

(0.686, 0.696)

–

Centralised

0.706

0.225

(0.701, 0.711)

1.10e−17

Local

0.659

0.220

(0.654, 0.665)

4.71e−05

AUROC

Distributed

0.723

0.182

(0.718, 0.727)

–

Centralised

0.729

0.180

(0.725, 0.734)

2.98e−26

Local

0.692

0.164

(0.688, 0.695)

2.48e−02

MAE

Distributed

2.370

1.608

(2.315, 2.425)

–

Centralised

2.365

1.923

(2.298, 2.431)

2.23e−04

Local

2.527

1.799

(2.465, 2.589)

9.01e−01

RMSE

Distributed

21.171

46.078

(19.584, 22.757)

–

Centralised

19.839

28.645

(18.853, 20.826)

2.92e−02

Local

23.771

49.776

(22.057, 25.485)

1.63e−01

  1. 2-sample T-test for the means was used as hypothesis test. Bold for P value below 0.05. AUPRC and AUROC for categorical target variable and RMSE and MAE for continuous target variable.
Back to article page