| Grant number: | 96/01741-7 |
| Support type: | Research Projects - Thematic Grants |
| Duration: | September 01, 1996 - August 31, 1999 |
| Field of knowledge: | Physical Sciences and Mathematics - Probability and Statistics |
| Principal Investigator: | Heleno Bolfarine |
| Grantee: | Heleno Bolfarine |
| Home Institution: | Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil |
| Associated grant(s): | 99/09157-0 - Pilar Loreto Iglesias Zuazola | Pontifical Catholic University of Chile - Chile,
AV.EXT 99/04541-7 - Theoperational Bayesian approach for finite populations, AR.EXT 98/03062-5 - Edward J. Stanek III | University of Massachusetts - Estados Unidos, AV.EXT + associated grants - associated grants |
Abstract
The statistical technique usually called regression analysis is mainly considered in situations where one is interested in studying the relationship between a set of characteristics (variables) considered as responses and another set of characteristics taken as potential explanatory factors. Typical examples may include the study of the relatioship between health expenditures and education, age and income or the investigation of the association between different risk factors such as smoking habits, obesity or hypertension and time to heart failure. Among the differente statistical models commonly considered to deal with such situations, we may highlight the so called measurement error models, longitudinal data models, survival analysis models and models for discrete data. The use of exact statistical inferential methods under these models is rarely possible and one must rely on approximations, the validity of which depends on the size of the available samples. In this context, the asymptotic methods considered in this project play a major role in the statistical analysis of the type of data with the structure described above. More specifically, the main goals of the project are: i) to develop inferential methodology for the different classes of regression models; ii) to optimize the statistical properties of such methods by means of appropriate large sample corrections; iii) to implement the proposed techniques computationally, and iv) to apply the results to problems of scientific, technological or artistical nature. (AU)