A non-negative spike-and-slab lasso generalized linear stacking prediction modeling method for high-dimensional omics data

Table 5 METABRIC data (N = 1420, events = 621)

	Deviance	AUC	BS	Misclassification
Penalty and group penalty methods
Lasso	909.438 (11.110)	0.677 (0.015)	0.225 (0.004)	0.374 (0.015)
MCP	921.284 (13.683)	0.662 (0.016)	0.229 (0.004)	0.385 (0.017)
SCAD	910.031 (10.286)	0.676 (0.014)	0.225 (0.003)	0.376 (0.015)
gsslasso	906.799 (12.149)	0.678 (0.014)	0.224 (0.004)	0.371 (0.014)
grlasso	909.698 (9.553)	0.682 (0.016)	0.225 (0.003)	0.368 (0.014)
grMCP	935.226 (146.534)	0.664 (0.020)	0.229 (0.012)	0.378 (0.017)
grSCAD	909.427 (10.034)	0.682 (0.016)	0.225 (0.003)	0.368 (0.015)
cMCP	951.428 (155.631)	0.662 (0.019)	0.233 (0.016)	0.381 (0.017)
Model stacking methods
nsslasso (Lasso)	907.198 (15.916)	0.683 (0.015)	0.224 (0.005)	0.369 (0.015)
nsslasso (network)	908.390 (15.794)	0.681 (0.015)	0.224 (0.005)	0.371 (0.014)
nsslasso (KNN)	951.428 (17.056)	0.660 (0.018)	0.229 (0.005)	0.383 (0.016)
nsslasso (NBa)	928.802 (13.804)	0.655 (0.016)	0.231 (0.005)	0.385 (0.015)
nsslasso (RF)	901.347 (12.669)	0.681 (0.014)	0.222 (0.004)	0.371 (0.013)
L-BFGS (Lasso)	905.644 (7.404)	0.688 (0.014)	0.223 (0.003)	0.371 (0.015)
nLasso (Lasso)	904.968 (14.092)	0.684 (0.015)	0.223 (0.004)	0.369 (0.014)

The prediction performance of stacking and the other methods (mean (SD)). Results are based on 100 random splits of the original data to training set (N = 710) and test set (N = 710) (21 candidate pathways)

ISSN: 1471-2105