Multivariate Growth Models

This thesis deals with models for (human) growth. We argue that growth is a multivariate phenomenon that should also be analysed as such. This is why we need multivariate growth models and multivariate reference curves.
An example of a multivariate growth model is the Multivariate Superimposition by Translation And Rotation model (MSITAR), an extension of the SITAR model by Cole. In the SITAR model, there is a general, or template function for the overall growth pattern which is modelled by a spline. From this template the individual growth curves are then derived by shifting it horizontally and vertically and stretching it in the horizontal direction.
Multivariate reference curves can be created by projecting the data in a large number of directions and then calculating the reference for that direction in the usual (univariate) manner. Quantile can be used to this end. Here we show that expectiles (are a generalisation of the mean in the same way that quantiles generalise the median) are an interesting alternative.