Finding Biomarker Signatures in Pooled Sample Designs: A Simulation Framework for Methodological Comparisons
Figure 4
Prediction errors for the second scenario: Ten linear patterns with noise levels of variance (a) and (b). On the x-axis the five statistical methods SVML, SVMR, RF, PPLS-DA and LDA after t-test are shown and on the y-axis the prediction error is given. The height of a bar specifies the average prediction error of 500 runs for a specific pool size and the 95% bootstrap confidence intervals are shown. The four bars for a method represent the different sizes of pools 1 (single samples), 2, 3 and 5 which are indicated by the numbers below the bars.
