Measuring Agreement

CHAPTER 5 REPEATED MEASUREMENTS DATA

5.1 PREVIEW

This chapter presents a methodology for the analysis of studies comparing two methods wherein both take multiple measurements on each subject. It focuses on two types of repeated measurements data—unlinked and linked. The usual steps in analysis consist of displaying data, modeling of data via a mixed-effects model, and associated evaluation of similarity and agreement. With repeated measurements, one can also evaluate repeatability of each method, which amounts to self-agreement. The methodology is illustrated with two case studies.

5.2 INTRODUCTION

Let Y_ijk denote the kth repeated measurement of the jth method on the ith subject. The data consist of Y_ijk, k = 1,...,m_ij, j = 1, 2, i = 1,...,n. Here n is the number of subjects in the study and m_ij is the number of measurements of method j on subject i. It is assumed that m_ij ≥ 2. Let M_i = m_i1 + m_i2 and , respectively, denote the total number of measurements on the ith subject and in the entire dataset. When m_i1 = m_i2, we will use m_i to denote the common value. Moreover, when the m_i are equal, the common value will be denoted by m. The design is balanced if each method has the same number m of measurements on every subject, otherwise it is unbalanced. Measurements on the same subject are dependent, whereas those from different subjects are assumed to be independent. It is necessary ...