Model Comparison Within Diagnostic Accuracy Test Model in Meta-Analysis

There many good of medicine knowledge, for example, therapy methods, effectiveness of new drug, association between disease and gene…etc., which didn，t get actually answers. We would need to find the really answers by collection the papers of related topic and combining the statistical results. This process is called systematic review in evidence-based medicine (EBM). Meta-analysis is one step of systematic review and is the statistical methods to pool some results within the collection papers to reply answers the EBM problems. Recently, meta-analysis is applied widely, especially in medicine, education, gene expression. Our approach will discuss the method of meta-analysis for diagnostic test. Diagnostic test for meta-analysis is very important in practice because for screen for disease or monitor treatment. This tool might a non-invasion test to classify a case as disease or nondisease by threshold. From 1993, many approaches were proposed in this field. So far there have not an adequate and coincident conclusion between these methods. Two of these methods, summary receiver operating characteristic (sROC) curve and bivariate model, are very common used in pooling results in diagnostic accuracy test. We would establish a criterion to select model between two methods to detect a proper model according data property. To get this criterion should through estimate generalized degree of freedom parameter (GDFP). We also extend to estimate multiple GDFP in this paper.