MEASURES OF AGREEMENT IN METHOD COMPARISON STUDIES FOR INTENSIVE LONGITUDINAL DATA

Open Access
Author:
Yoo, Mina
Graduate Program:
Statistics
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
May 05, 2011
Committee Members:
  • Dr Runze Li, Dissertation Advisor
  • Runze Li, Committee Chair
  • Mosuk Chow, Committee Chair
  • Naomi S Altman, Committee Member
  • Vernon Michael Chinchilli, Committee Member
Keywords:
  • concordance correlation coefficient
  • varying coefficient models
  • partially linear models
  • mixed effects models
  • measurement agreement
  • intensive longitudinal data
Abstract:
This dissertation is concerned with assessment of measurement agreement for intensive longitudinal data. Assessment of the measurement agreement encompasses a variety of applications. A number of indices for measuring agreement have been developed. However, these measures make a major assumption: that the mean and variation are stable over time. With recent developments in data collection methods and statistical models, intensive longitudinal studies and the analysis of intensive longitudinal data are gaining popularity across many areas. Intensive longitudinal data enable researchers to examine more detailed features of how pro- cesses change over time. Due to its high intensity of assessments within subjects, it has different characteristics from traditional longitudinal data, which often involve a small number of repeated observations across many individuals. The overall mean of intensive longitudinal data is typically a smooth curve of time and variance of the error process may be time-varying over study duration. Moreover, heterogeneity of intra subject processes such as autocorrelation and instability exists. To overcome these challenges and provide accuracy estimates, we first propose a novel estimation procedure for functional mixed models and partially linear mixed models and study the asymptotic properties of the proposed estimation procedure. Then, we develop a new index of the agreement for intensive longitudinal data, the functional type of concordance correlation coefficient based on proposed models. The functional concordance correlation coefficient is robust with respect to model specification, compared with the popular index, the unified approach of concordance correlation coefficient. The proposed index improves the accuracy of measurement agreement by separating the time trend of measurements from the degree of agreement. All the proposed procedures are assessed by intensive finite sample simulation studies and most are illustrated with real data examples.