Estimation and hypothesis test for parameters of binary diagnostic tests

Abstract

The likelihood ratios of a binary diagnostic test are very widely used parameters to compare the effectiveness of two binary diagnostic tests, as they are technically equivalent to a relative risk. This Doctoral Thesis proposes new confidence intervals to compare the positive (negative) likelihood ratios of two binary diagnostic tests subject to a paired design, analysing the problem both from a frequentist perspective and a Bayesian one. The simulation experiments carried out to compare the asymptotic behaviour of the confidence intervals have allowed us to give some general rules of application for the intervals studied. Furthermore, a method is proposed to calculate the sample size to compare the positive (negative) likelihood ratios of two binary diagnostic tests through a confidence interval.