Nonparametric and semiparametric methods in medical diagnosticsPublic Deposited
Add to collection
You do not have access to any existing collections. You may create a new collection.
Downloadable ContentDownload PDF
MLAKim, Eunhee. Nonparametric and Semiparametric Methods In Medical Diagnostics. Chapel Hill, NC: University of North Carolina at Chapel Hill, 2009. https://doi.org/10.17615/0316-tb89
APAKim, E. (2009). Nonparametric and semiparametric methods in medical diagnostics. Chapel Hill, NC: University of North Carolina at Chapel Hill. https://doi.org/10.17615/0316-tb89
ChicagoKim, Eunhee. 2009. Nonparametric and Semiparametric Methods In Medical Diagnostics. Chapel Hill, NC: University of North Carolina at Chapel Hill. https://doi.org/10.17615/0316-tb89
- Last Modified
- March 21, 2019
- Affiliation: Gillings School of Global Public Health, Department of Biostatistics
- In medical diagnostics, biomarkers are used as the basis for detecting or predicting disease. There has been an increased interest in using the Receiver Operating Characteristic (ROC) curve to assess the accuracy of biomarkers. In many situations, a single biomarker is not sufficient for the desired level of accuracy; furthermore, newly discovered biomarkers can provide additional information for a specific disease. Even though numerous methods have been developed to evaluate a single biomarker, few statistical methods exist to accommodate multiple biomarkers simultaneously. The first paper proposes a semiparametric transformation model for multiple biomarkers in ROC analysis to optimize classification accuracy. This model assumes that some unknown and marker-specific transformations of biomarkers follow a multivariate normal distribution; it incorporates random effects to account for within-subject correlation among biomarkers. Nonparametric maximum likelihood estimation is used for inference, and the parameter estimators are shown to be asymptotically normal and semiparametrically efficient. The proposed method is applied to analyze brain tumor imaging data and prostate cancer data. In the second paper, we focus on assessing the accuracy of biomarkers by adjusting for covariates that can influence the performance of biomarkers. Therefore, we develop an accelerated ROC model in which the effect of covariates relates to rescaling the original ROC curve. The proposed model generalizes the usual accelerated failure time model in the survival context to the ROC analysis. An innovative method is developed to construct estimating equations for parameter estimation. The bootstrapping method is used for inference, and the parameter estimators are shown to be asymptotically normal. We apply the proposed method to data from a prostate cancer study. The paired-reader, paired-patient design is commonly used in reader studies when evaluating the diagnostic performance of radiological imaging systems. In this design, multiple readers interpret all test results of patients who undergo multiple diagnostic tests under study. In the third paper, we develop a method to estimate and compare accuracies of diagnostic tests in a paired-reader, paired-patient design by introducing a latent model for test results. The asymptotic property of the proposed test statistics is derived based on the theory of U-statistics. Furthermore, a method for correcting an imperfect gold standard bias and sample size formula are presented. The proposed method is applied to comparing the diagnostic performance of digital mammography and screen-film mammography in discriminating breast tumors.
- Date of publication
- December 2009
- Resource type
- Rights statement
- In Copyright
- "... in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Biostatistics."
- Ibrahim, Joseph
- Zeng, Donglin
- Place of publication
- Chapel Hill, NC
- Access right
- Open access
- Date uploaded
- March 18, 2013
This work has no parents.
|Nonparametric and semiparametric methods in medical diagnostics||2019-04-11||Public||Download|