Bayesian compositional regression with microbiome features via variational inference

BMC Bioinformatics

Table 4 Table of true positive rate, false positive rate, L2 loss and prediction error for the additive-log-ratio model with SNR of 0.83 and d of 200

Method	\(\rho\)	TPR	FPR	L2	PE
Lasso	0	0.965 ± 0.068	0.120 ± 0.084	10.238 ± 2.330	1.998 ± 1.127
Lin	0	0.962 ± 0.082	0.135 ± 0.089	10.070 ±2.661	2.002 ± 1.114
Bates	0	0.887 ± 0.141	0.004 ± 0.009	11.923 ± 1.318	1.909 ± 1.196
Selbal	0	0.687 ± 0.150	0.000 ± 0.000	14.837 ± 1.199	2.424 ± 1.475
VB	0	0.725 ± 0.104	0.000 ± 0.001	12.706 ± 1.119	1.962 ± 1.061
Lasso	0.2	0.972 ± 0.075	0.134 ± 0.076	9.911 ± 2.239	2.045 ± 1.214
Lin	0.2	0.972 ± 0.071	0.158 ± 0.079	9.580 ± 2.344	2.351 ± 1.443
Bates	0.2	0.925 ± 0.126	0.004 ± 0.009	11.691 ± 1.368	1.806 ± 1.178
Selbal	0.2	0.657 ± 0.074	0.001 ± 0.001	14.494 ± 1.011	2.649 ± 1.624
VB	0.2	0.720 ± 0.170	0.000 ± 0.001	12.947 ± 1.908	2.186 ± 1.496
Lasso	0.4	0.847 ± 0.125	0.142 ± 0.084	10.054 ± 2.316	3.142 ± 1.834
Lin	0.4	0.810 ± 0.121	0.172 ± 0.084	9.630 ± 2.456	3.341 ± 1.861
Bates	0.4	0.798 ± 0.142	0.008 ± 0.015	11.790 ± 1.642	1.860 ± 1.116
Selbal	0.4	0.633 ± 0.085	0.001 ± 0.001	13.986 ± 1.097	2.669 ± 1.896
VB	0.4	0.603 ± 0.139	0.000 ± 0.001	13.611 ± 1.971	2.418 ± 1.797

ISSN: 1471-2105