Figure 4
From: Early symptoms and sensations as predictors of lung cancer: a machine learning multivariate model
Orthogonal projections to latent structures (OPLS) 3D scores plot. Individual scores for the training set (A n = 433) and predicted scores (PS) for the cross-validated test set (B n = 73) are shown for the final model. All three of the OPLS model components are plotted, including the predictive component (t[1]) and the two orthogonal components (to[1] & [2]) (total R2X variance = 42.3%: t[1] = 23.6%, to[1] = 12.7%, to[2] = 6%). Predictive explained R2Y variance (lung cancer: training set): 62.4%; cross-validated explained Q2 variance (lung cancer: cross-validated test set): 58.1%. A total of 63 descriptors of symptoms and sensations were included together with seven background variables (Table 2). Coloured circles indicate lung cancer (red) or no cancer (blue). Outliers are indicated beyond the 95% confidence interval ellipse.
