Skip to main content

Table 1 Table of true positive rate, false positive rate, L2 loss and prediction error for the additive-log-ratio model with SNR of 0.5 and d of 45

From: Bayesian compositional regression with microbiome features via variational inference

Method

\(\rho\)

TPR

FPR

L2

PE

Lasso

0

0.937 ± 0.091

0.262 ± 0.136

18.465 ± 1.933

4.741 ± 2.901

Lin

0

0.972 ± 0.067

0.266 ± 0.124

18.572 ± 1.847

4.828 ± 2.928

Bates

0

0.822 ± 0.147

0.029 ± 0.048

19.504 ± 1.803

4.837 ± 3.059

Selbal

0

0.655 ± 0.117

0.002 ± 0.008

21.159 ± 1.541

5.305 ± 3.570

VB

0

0.678 ± 0.111

0.002 ± 0.008

20.072 ± 1.560

4.687 ± 2.948

Lasso

0.2

0.882 ± 0.133

0.271 ± 0.140

18.442 ± 1.996

5.051 ± 2.951

Lin

0.2

0.897 ± 0.125

0.292 ± 0.143

18.526 ± 2.215

4.939 ± 2.866

Bates

0.2

0.767 ± 0.128

0.042 ± 0.057

19.425 ± 1.828

5.023 ± 2.727

Selbal

0.2

0.637 ± 0.114

0.005 ± 0.012

21.259 ± 1.718

4.815 ± 2.800

VB

0.2

0.640 ± 0.141

0.002 ± 0.007

20.156 ± 1.549

4.410 ± 2.475

Lasso

0.4

0.825 ± 0.137

0.294 ± 0.163

18.330 ± 2.118

4.661 ± 3.396

Lin

0.4

0.835 ± 0.139

0.354 ± 0.151

18.322 ± 2.180

4.651 ± 3.205

Bates

0.4

0.692 ± 0.165

0.058 ± 0.079

19.471 ± 2.169

4.583 ± 3.437

Selbal

0.4

0.550 ± 0.168

0.004 ± 0.012

21.134 ± 1.699

4.733 ± 3.852

VB

0.4

0.428 ± 0.211

0.001 ± 0.005

20.835 ± 2.021

4.370 ± 2.948

  1. Bates refers to the two-stage log-ratio lasso and VB refers to the CAVI-MC