ABSTRACT

In clinical trials involving measurements of biomarkers, the values supplied are typically estimates and hence subject to measurement errors (MEs). This chapter reviews different methods to operate with data subject to various sorts of MEs problems. It presents a limited number of Monte Carlo results to compare the efficiency and show the applicability of the considered methods. The chapter also presents several methods for evaluating data subject to additive errors or errors caused by limits of detection (LODs) problems. It considers parametric and nonparametric inference based on data obtained following the repeated measurements design. The chapter shows that the likelihood methodology can be easily applied to develop statistical procedures based on data subject to additive errors. It describes a novel hybrid design that requires a consideration of assays on individual specimens and assays on pooled specimens when the measurement process is subject to the LODs problem.