Skip to main content

Table 2 Posterior predictive comparisons on simulated data

From: Comparing Bayesian hierarchical meta-regression methods and evaluating the influence of priors for evaluations of surrogate endpoints on heterogeneous collections of clinical trials

 

FP-RE

NP-RE

PP-RE

 

PP-RE

PP-FE

  
 

Cvg

Cvg

Cvg

RR\(_{np}\)

WR\(_{np}\)

 

Cvg

Cvg

RR\(_{fe}\)

WR\(_{fe}\)

Setup 1 (V1)

Setup 1 (V2)

SG 1

0.99

1.00

1.00

1.813

2.685

SG 1

1.00

0.97

1.352

0.861

SG 2

0.99

1.00

1.00

1.549

3.027

SG 2

1.00

0.97

1.319

0.844

SG 3

0.99

1.00

1.00

1.773

3.246

SG 3

1.00

0.97

1.362

0.857

Setup 2 (V1)

Setup 2 (V2)

SG 1

0.87

0.95

0.90

2.035

2.642

SG 1

0.90

0.90

0.912

1.041

SG 2

0.94

0.96

0.94

2.250

2.766

SG 2

0.94

0.93

0.937

1.078

SG 3

0.97

0.99

0.99

2.234

2.600

SG 3

0.99

0.91

1.584

1.057

Setup 3 (V1)

Setup 3 (V2)

SG 1

0.79

0.95

0.89

1.892

2.492

SG 1

0.89

0.90

0.887

1.000

SG 2

1.00

1.00

0.99

1.772

2.668

SG 2

0.99

0.95

1.529

1.090

SG 3

0.99

0.99

0.99

1.925

2.726

SG 3

0.99

0.92

1.670

1.057

Setup 4 (V1)

Setup 4 (V2)

SG 1

0.92

0.96

0.92

1.886

2.502

SG 1

0.92

0.91

0.923

1.116

SG 2

0.89

0.94

0.90

1.888

2.630

SG 2

0.90

0.90

0.881

1.046

SG 3

0.98

0.99

0.99

1.697

2.507

SG 3

0.99

0.92

1.592

1.112

  1. V1 Setups where true surrogate effects are Gaussian, V2 Setups with Non-Gaussian true surrogate effects, SG “Subgroup.”, Cvg Coverage, RR\(_{np}\) Ratio of NP-RE prediction RMSE over PP-RE prediction RMSE, WR\(_{np}\): Ratio of NP-RE average 95% PPD width to PP-RE average 95% PPD width, RR\(_{fe}\) Ratio of PP-FE prediction RMSE over PP-RE prediction RMSE, WR\(_{fe}\) Ratio of PP-FE average 95% PPD width to PP-RE average 95% PPD width