Measuring Agreement: Methodology and Applications

This book deals with the topic of statistical evaluation of agreement among two or more methods of measurement of a variable.  The primary focus is on a continuous variable.  In practice, a measurement method may be a medical device, an instrument, an assay, a clinical observer, a questionnaire, or a technique or technology. The variable being measured typically has clinical importance, i.e., blood pressure, heart rate, cholesterol level, etc., and readings on individuals form the basis of their health assessment.  As technology advances, new measurement methods become available that may be cheaper, less invasive, or more reliable than older methods.  However, before a new method is adopted for general use, it is validated by comparing it with an established method in a method comparison study.  The basic statistical problem in a method comparison study is to quantify the extent of agreement between the measurement methods and determine if it is sufficient.   The book implements the methodologies described throughout using the statistical software R, and a related Web site contains the datasets and R software code for all of the case studies.  Chapter coverage includes measures of agreement for continuous response and categorical responses; likelihood-based inference; Bayesian inference; nonparametric and semiparametric inference; censored measurements; regression methods for incorporating covariates; sample size determination; comparison of multiple measurement methods; and measurement error models.