Lambert Paul C, Burton Paul R, Abrams Keith R, Brooke Adrian M
Centre for Biostatistics and Genetic Epidemiology, Department of Health Sciences, University of Leicester, 22-28 Princess Road West, Leicester LE1 6TP, UK.
Stat Med. 2004 Dec 30;23(24):3821-39. doi: 10.1002/sim.1951.
Peak expiratory flow (PEF) is a measure commonly used in clinical practice and research for respiratory diseases such as asthma. In research, PEF is usually recorded in a diary for a 2-week period with two or more measurements per day. Interest may lie in whether certain groups of individuals tend to have higher or lower PEF. In addition the variability of PEF may be of interest as, for example, asthmatics tend to have more variable airways. In this paper we develop a three-level hierarchical model that can simultaneously model the mean level and variability of PEF. The variability is broken down into three components, between-subject variability, between-day within-subject variability, and within-day within-subject variability. The latter two components are of specific clinical interest. We fit both classical and Bayesian models. The Bayesian models have the advantage of taking the uncertainty in the variance component estimates into account when estimating the standard errors of the fixed effects. In addition, the Bayesian models provide an intuitive and simple way to investigate the within-subject variance components.
呼气峰值流速（PEF）是临床实践和哮喘等呼吸道疾病研究中常用的一项指标。在研究中，PEF通常会在日记中记录两周时间，每天测量两次或更多次。研究兴趣可能在于某些个体群体的PEF是否往往较高或较低。此外，PEF的变异性也可能受到关注，例如，哮喘患者的气道往往具有更大的变异性。在本文中，我们开发了一种三级分层模型，该模型可以同时对PEF的平均水平和变异性进行建模。变异性被分解为三个组成部分，即个体间变异性、个体内日间变异性和个体内日内变异性。后两个组成部分具有特定的临床研究意义。我们拟合了经典模型和贝叶斯模型。贝叶斯模型的优势在于，在估计固定效应的标准误差时，会考虑方差分量估计中的不确定性。此外，贝叶斯模型提供了一种直观且简单的方法来研究个体内方差分量。
