Título: Exploring reliability in epidemiology and clinical research
Autores: Rodella, Stefania
Fecha: 1996
Publicador: McGill University - MCGILL
Fuente:
Tipo: Electronic Thesis or Dissertation
Tema: Biology, Biostatistics.
Health Sciences, General.
Descripción: Quality of measurement is a fundamental issue in medical research and clinical practice, of relevance for any inference or decision-making process. Reliability, as an estimate of the variability in the domain of repeated measurements, is an important component of the quality of measurements.
Reliability indices are frequently used and presented in the medical literature and a considerable amount of methodological research has been conducted on this topic in the last decades. However, the debate is still open on some theoretical and operational aspects. Available knowledge concerning reliability, particularly for categorical data, is not easily accessible since it is often confined to specialized journals and almost disregarded by statistical textbooks. Therefore, a thorough understanding is difficult to achieve for a researcher potentially involved in reliability studies.
My main objective was to pursue a conceptual and global understanding of the role of reliability in the domain of categorical data. In order to achieve this goal I reviewed and synthesized the literature according to some specific objectives: (a) to provide an overview on the founding concepts and methods in the measurement of reliability for categorical variables, also contrasting them with what has been done in the domain of continuous variables; (b) to present and discuss the main limitations of traditional indices, particularly the kappa statistic; (c) to briefly introduce some possible alternative methods and areas for future development; (d) to emphasize the implications of reliability for epidemiological and clinical research.
Finally, in order to illustrate the application of some of the methods discussed, I used a real set of data, concerning 209 slides of lymphomas tissue samples, reviewed by a panel of four pathologists, according to a standard classification based on 10 categories.
Idioma: en