Figure 2
From: Probabilistic principal component analysis for metabolomic data
Results of fitting a PPCCA model to the urine dataset with weight as a covariate. A. Plot of the modified BIC values and the proportion of variation explained by each model: the higher the BIC value the better the model. B. The scores plot: the red dots denote the subjects in the treatment group and the black dots denote those in the control group. Dot size reflects a subject's weight. (Larger dots suggest heavier subjects.) The grey ellipses illustrate 95% posterior sets indicating the uncertainty associated with each score. C. A barplot of spectral bins with loadings which are significantly different from zero and greater in absolute value than 0.8. The barplot shows how the selected spectral regions load on PC 1 and their corresponding 95% confidence intervals.