Evaluation of diagnostic tests when there is/is not A gold standard: Contribution of bayesian approach

Diagnostic tests are used for defining the disease status of person in a heterogeneous population which consists of healthy and diseased people. With the help of the gold standard tests one can be diagnosed as "diseased" or "healthy", however these tests may not be used in each uncertain situation due to the difficulty on practice, high costs and being interventional in some diseases. For this reason in many fields of medicine, diagnostic tests are developed to be used in place of gold standard tests. Measures of accuracy that show the power of classification of the newly developed alternative tests are obtained by using information provided by the gold standard test. In the absence/presence of a gold standard test these measures of accuracy can be determined using different approaches, one of which is Bayesian approach commonly used by the researchers in many different fields. The Bayesian approach also provides a way to include expert prior knowledge concerning parameters of interest. Nowadays the calculation of probability on clinical science is using on nomograms etc. and the contribution is undeniable. The aim of this thesis is to evaluate diagnostic tests by using Bayesian approach whether or not there is a gold standard, to assess the contribution of Bayesian approach, to analyze the datasets for practical application with WinBUGS and/or R package programme. Datasets will be analyzed with WinBUGS package programme. As statistical methods, Bayesian inference, Gibbs sampling and Markov Chain Monte Carlo will be employed. Diagnostic test for a disease give more reliable results when there is a Bayesian inference that including prevalence information.