Figure 4
From: Multivariate meta-analysis of proteomics data from human prostate and colon tumours
Prediction success for different PLS model parameter settings. Average of geometric mean of sensitivity and specificity over 500 bootstrap rounds in inner loop using different number of PLS components and different number of variables. Results for missing value method "ten lowest". Black lines are results for the VIP selected variables and grey lines are the corresponding results for the same number of randomly picked variables. Solid lines represent two PLS components, dashed three PLS components and dotted four PLS components.