Abstract:
|
Evaluate the accuracy (sensitivity and specificity) of new diagnostic test without the presence of a gold standard is of practical meaning and has been the subject of intensive study for several decades. Existing methods use two or more diagnostic tests with a latent model and with the accuracy implemented as parameters, then estimate them via the maximum likelihood estimate (MLE). Various variations of this method and improvements of this model under some conditions are proposed in the literature. However, the existing method requires two populations to evaluate two diagnostic tests, assumes that the two tests are independent conditional on the disease statuses which is unreasonable in many situation. Although many methods to accommodate dependence between the two tests have been developed, these methods further complicate the model structure and made further assumptions for a specified dependence model. To evaluate the accuracy of new diagnostic test without the dependence assumption, we propose to first estimate independent functions of the accuracy parameters by empirical method, then solve for these accuracy parameters from these functions. The proposed method is automatically
|