|
Abstract:
|
Diagnostic biomarkers play a critical role in biomedical research such as diagnosis and prediction of diseases, etc. To improve the diagnostic performance, considerable research about combining multiple biomarkers have been derived based on the multivariate normality assumption. But it is common in practice that most biomarkers follow distributions that are far from normality. In my project, I focus on the combination of elliptical distribution which is a broader distribution family. It is well known that the likelihood ratio combination is the optimal combination but it’s complicated to calculate. In my study, the ROC curve function of elliptical likelihood ratio (ELR) combination is derived. Based on the nonparametric maximum likelihood estimate (NPMLE), I build the empirical estimation of ELR combination. Simulation results show that my method could obtain the optimal combination. The proposed method is applied to the diagnosis of neural tube defects (NTD). The elliptical likelihood ratio combination improves the AUC value by 11.43% and 23.29% from normal likelihood ratio combination and empirical best linear combination, respectively.
|
