In this project we discuss the development and analysis of nonlinear elliptical models, which represents a new methodology for obtaining robust estimates to nonlinear correlated data. The nonlinear models have been widely used, but little can be found using the elliptical approach. We intend to consider models which were previously analysed under normality, but that presented heavy tail when compared to the gaussian distribution, and reanalyse them using elliptical errors, in order to obtain better fit and robustness. We also intend to develop sensitivity tools, based on the residuals analysis and local influence, which can help on validation and model choice. As numerical illustrations, we will analyse real data, like kinetics data or growth curves, and develop computational tools for the necessary analysis. As results, beyond the doctoral thesis, we expect to achieve publications on important international journals and scientific meetings.
News published in Agência FAPESP Newsletter about the scholarship: