WITHDRAWN: Generalized Additive Models to Nomial Responses Using Bivariate Kernel: A Solution to Spatial Analysis

• Author(s): Ana Carolina Cintra Nunes Mafra and Luciana Bertoldi Nucci and Ricardo Carlos Cordeiro and Celso Stephan
• Keywords: generalized additive models ; multinomial response ; bivariate Kernel

Abstract

Spatial analysis are frequently used in case-control studies in a binomial format, however many studies have more than one category of cases. Thus, the best analysis need to include this information and a multinomial analysis is proposed here. Generalized additive models are extended to obtain a defined spatial risk. The forms to analyze and map it through them, including acquaintance of the estimates significance, were presented. As an ilustration a data from a case-control study of occupational accidents was used with three categories: severe cases, mild cases and controls. The analysis has found areas of increased risk that varied depending on the level of comparison. In parametric variables studied different risk and protection factors were found. This brings the ability to analyze data from case-control studies with multinomial response where the spatial risk has to be analyzed.