IV. RESULTS
A. Detecting presence of PD: binary classification task
For the two cases of equal misclassification costs for PD
and control data points, and varying misclassification costs
where the cost of misclassifying a control subject as PD is
assigned a 50% higher cost compared to misclassifying a PD
subject as control, the prediction performance is given in
Table II. Sensitivity measures how well the prediction model
recognizes PD cases. Specificity measures how well the
model identifies control cases. The false positive rate
measures how often controls are misclassified as PD.
Precision, or positive predictive value, measures how well
the prediction of PD reflects the underlying presence of
disease.