1) Partial Functional Area under the Curve Regression: a Metabolic Syndrome Case Study – 2) Spectral Field- Based Modeling for Bivariate Extremes


  • Profª Vanda Inácio e Prof. Miguel Carvalho – Pontificia Universidad Católica de Chile
  • FCUL (DEIO) – Campo Grande – Bloco C6 Piso 4 – Sala 6.4.30 – 14:00h às 16:00h
  • Terça-feira, 10 de Dezembro de 2013
  • Referência Projeto: PEst-OE/MAT/UI0006/2011
Seminário 1

Abstract: The statistical evaluation of diagnostic tests is of great importance in public health and medical research. New diagnostic tests must be rigorously evaluated to determine their abilities to discriminate between diseased and nondiseased states and it is of crucial importance to understand the covariate influence to determine the optimal and suboptimal populations to perform such tests on. Due to advances in technology, medical diagnostic data have become increasingly complex and, nowadays, applications where measurements are functions, curves, or images are becoming more and more common. We develop nonparametric regression methods for the partial area under the receiver operating characteristic curve, a well-accepted measure of diagnostic test accuracy when only a region of the curve is of interest, for the case where the covariate  influencing test’s performance is functional. The simulation study shows that our method produces estimates with small bias and small mean squared error. The application of our method to assess the ability of the gamma-glutamyl-transferase to detect women with metabolic syndrome reveals that the nocturnal levels of arterial oxygen saturation of hemoglobin are key to test performance. Joint work with Miguel de Carvalho, Todd A. Alonzo and Wenceslao Gonzalez-Manteiga.

Seminário 2

Abstract: The spectral density is a mainstay of bivariate extreme value modeling, being particularly relevant for assessing extremal dependence. In this talk I introduce a regression model for the spectral density of a bivariate extreme value distribution that allows us to assess how extremal dependence evolves over a certain covariate. For estimating our model we propose a double kernel estimator, which can be seen as an extension of the Nadaraya–Watson estimator where the usual scalar responses are replaced by mean-constrained densities on the unit simplex. A case study is used to illustrate our methods. This is joint work with D. Castro and J. Wadsworth.