Additional file 2 of Physiological serum uric acid concentrations correlate with arterial stiffness in a sex-dependent manner

Additional file 2. Additional information on the machine learning procedure. Table. S01: Percent increaseof mean squared errorfor each of the 38 variables used to predict PWV with random forests in a 10-fold CV scheme. Hypertension = measured systolic blood pressure ≥ 140 or diastolic blood pressure ≥ 90; hypertension= self-reported hypertension; cardiovascular diseases = intermittent claudication or circulatory disorders in the legs, also known as intermittent claudication or arterial occlusive disease; 1st cancer type = entity of the first diagnosed cancer if cancer has ever been diagnosed. Fig. S02: Boxplots for each individual fold. These results indicate how well the 10 individual random forestsdid adapt to the non-linear regression task and to what extent results varied across all ten folds. Red lines indicate a threshold of +5% or −-5%, respectively. Table. S02: Descriptive statistics for each individual fold. These results equal the data visualized in Fig. S02. A mean error of 1% means that a given test data sample, the random forest did on average deviate from the actual target value by 1% of that value