Title: Combination of multiple functional markers to improve diagnostic accuracy
Authors: Qinyi Zhang - The Hong Kong Polytechnique University (Hong Kong) [presenting]
Abstract: Powerful diagnostic marker is important in diagnosis. Because of development of modern technique, diagnostic markers can be observed repeatedly and act as functional markers. Existing methods mainly discussed diagnosis by a single scalar marker or combinations of multiple scalar markers but not functional markers. Methods of functional data analysis are used to make diagnosis for functional markers. In particular, we adopt functional principal components analysis to obtain basis functions and the corresponding projections, derive the features on the basis of the projections, and finally the combinations of the obtained features. Receiver operating characteristic (ROC) curve is widely used for evaluating diagnosis, which can be assessed by area under the curve (AUC) or Youden Index. Our proposed methods are illustrated by simulations and real data analysis of diagnosis for high- or low- hospital admissions due to respiratory diseases in Hong Kong.