| A class of nonparametric models to describe the influence of covariates on curve data is introduced. This approach is based on the concept that each observed curve is the realization of a random process. A functional random effects model is proposed to allow the covariate to act through the stochastic terms of a principal component expansion. Smoothing methods are used for the estimation of the functional fixed and random components of these models and a basic consistency result is obtained. The proposed models are illustrated in an analysis of a data set consisting of egg-laying curves for 1000 female Mediterranean fruit flies. In this application, the proposed model is seen to be useful in relating the shapes of the age-specific fecundity curves to the total number of eggs laid. (This is a joint work with Hans-Georg Mueller and Jane-Ling Wang.) |