Figure 1From: Integrating multi-platform genomic data using hierarchical Bayesian relevance vector machinesKernel density estimates. Kernel density estimates (KDEs) of mean square prediction errors (MSPEs) for H-RVM, Gene-Lasso, MiRNA-Lasso, and Interaction-Lasso. The GBM survival data is randomly split 50 times into training and test data, all four models are fit on the training data, and MSPEs for log survival times are calculated for the test data using the model fit on training data. The x-axis represents the MSPEs and the y-axis represents the respective KDEs for H-RVM (in solid red), Gene-Lasso (in dotted blue), MiRNA-Lasso (in dotted and dashed blue), and Interaction-Lasso (in dashed blue).Back to article page