Abstract: This paper describes methods for modeling the dependence of the level and rate of growth of pulmonary function during childhood on two physiologic variables, age and height, and for assessing the effects of individual and environmental risk factors on measures of pulmonary function. Descriptive analyses stratified by age suggest that the relation between pulmonary function and height in children is linear but age-dependent. Thus, we consider models in which pulmonary function level (or rate of growth) depends linearly on height (or change in height), but with an agedependent intercept and slope. Regression splines are used to describe the change in intercept and slope with age. To accommodate repeated measures and heterogeneity of variance, robust variance estimates are derived for the estimated regression coefficients. The methods presented provide a flexible family of growth curves, quantify the effects of covariates on level and rate of growth, and have attractive clinical and epidemiological interpretations.
Key words and phrases: Growth curve models, regression splines, successive differences, robust variance estimation, pulmonary function growth.