Generalized Additive Models for Location, Scale and Shape (GAMLSS) for Modeling Reference

Abstract

Reference curves for values of anthropometric parameters of a population are widely used by healthcare professionals. Statistical techniques or the construction of age-related mathematical models have developed a lot in the last 10 years and this has been an area of particular interest in several research centers around the world. The evolution of the processing power of personal computers has contributed significantly to this change in the methods of regression analysis used until then. The WHO publication of the Multicenter Growth Reference Study (MGRS) in 2006 created a new paradigm for the construction of growth curves, which, consequently, resulted in a broad discussion on the curves previously published and currently in use in several healthcare areas. The aim of this exhibit was to expose in a concise way — without further consideration of the mathematical aspects — the methodological bases of statistical modeling using a new class of statistical models called Generalized Additive Models for Position, Scale and Shape (GAMLSS) proposed by RIGBY and STASINOPOULOS (2005). The application of the GAMLSS method to design a model with empirical data was demonstrated, as well as the use of diagnostic tools and the selection of the model of best fit, the construction of a curve of reference and easier calculation of percentiles and z-scores through specific functions. It is concluded that the GAMLSS method is an important tool for statistical modeling in healthcare and that in future research it will probably be the state of the art in the construction of new reference curves or in the validation of curves already built. Epidemiological Methods. Statistical Models. Regression. Analysis