Figure 2From: A systematic evaluation of high-dimensional, ensemble-based regression for exploring large model spaces in microbiome analyses Receiver operating characteristic (ROC) curves. Mean ROC curves across 130 simulations are shown for approaches that do not perform variable selection. An ideal ROC has a high TPR and a low FPR over a wide range of thresholds. S denotes the sparsity setting or the proportion of influential variables in the data. The diagonal dashed line represents random variable selection performance. a. β=±1 and b. \({\beta \in \mathcal {U}}\).Back to article page