Marginal and dynamic regression models for longitudinal data.

摘要: Regression models for longitudinal data often employ random effects and serial correlation to account residual variation between within subjects. Most of these are marginal models, separating the mean covariance parameters. This paper discusses use dynamic in which a lagged response serves as predictor compares models. parameters have different interpretation they describe changes levels, rather than levels themselves. Lagged predictors also useful with data, explicitly quantifying effect previous risk factors. These explored through analysis from Childhood Respiratory Study, modelling lung function (FEV1) age, height, sex smoking status children measured over five-year period. Copyright © 2001 John Wiley & Sons, Ltd.