Longitudinal principal components and non-linear regression models of early childhood growth.

The Jenss curve is a non-linear regression model which describes quite well the longitudinal length or weight measurements of an individual child from three months to six years. It is contrasted with modelling growth by longitudinal principal components analysis, an approach developed by Kent which is not restricted to any particular curve type or age range, and which can be thought of as providing the best fit among linear models having kappa parameters. Thus, in addition to being a model itself, it can be used to evaluate the success of other models. It is shown how this method differs from classical principal components. Data from Longitudinal Studies of Child Health and Development are used to explore these issues.