Year: 1998
Type: editorial
Abstract: Variance components in medical research`That haphazard variability may have quite complex structure and thus not be represented mathematically by independent and identically distributed random variables is important in many areas of application.In that sense the splitting of variation into components and in particular the study of variance components can be claimed to be a central theme in statistics with wide rami®cations in both applications and theory.'This is the opening paragraph of Sir David Cox's review of historical developments in variance components, the ®rst of ®ve outstanding papers on variance components in medical research to which this issue of the journal is devoted.Components of variance are an old idea enjoying a renaissance of interest.Classically, they have been regarded as essential for the proper study of measurement error, of reliability in industrial processes, for attributing sources of variability in genetics particularly for animal and plant-breeding experiments, and more.Past students of statistics may well remember grappling with split-plots and other multistratum analysis of variance models!More recently, attention has focused on linear and nonlinear random effects models especially generalized linear mixed models and hierarchical models.Random effects and their associated variance components have been central, for instance, to efforts to successfully model markers of HIV/AIDS disease progression.That variance components are moving to centre stage once again is partly due to the easy availability of computing power and appropriate statistical software, but also to the recognition that partitioning observed variability into its component parts is essential for valid statistical analysis and interpretation.The papers presented here re¯ect this modern emphasis in modelling and inference for complex random phenomena.Sir David Cox sets the scene with a timely review of early work on variance components, some of which is in danger of being overlooked.He establishes many useful connections between this earlier work and current directions, and sets out the general principles and concepts involved in considering variance component models.Sir David also gives some recommendations for handling nonstandard problems, model checking, unbalanced data and outliers, and gives us a clear notion of what variance components are really all about.There is a history of applications of variance components to the analysis of data on blood pressure which is characterized by its variability; see for instance some early work by Armitage and Rose.1Y2 Over the past decade there has been intense interest in modelling variability in clinical markers for HIV infection and AIDS, notably CD4 cell counts and more recently for other markers such as viral load.This is the subject of the second paper in this issue by John Boscardin, Jeremy Taylor and Ngayee Law.These authors present a review of recent approaches to the modelling and analysis of longitudinal marker data, ranging from simple random coef®cient models for individual CD4 cell count trajectories to models incorporating more technical Ó
Author: Patty Solomon
Institution: The University of Adelaide
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4240032390
