Full text
| |
| Author(s): |
Áurea Fonseca Lopes Galindo
Total Authors: 1
|
| Document type: | Master's Dissertation |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Instituto de Matemática, Estatística e Computação Científica |
| Defense date: | 2021-03-01 |
| Examining board members: |
Caio Lucidius Naberezny Azevedo;
Fernanda de Bastiani;
Guilherme Vieira Nunes Ludwig
|
| Advisor: | Juvêncio Santos Nobre; Caio Lucidius Naberezny Azevedo |
| Abstract | |
The main focus of this work consists on the modeling of (in)dependent data in the interval [0,1], and the respective particular cases. In this sense, we developed two approaches, under the frequentist paradigm. The first one corresponds to a new class of models based on the beta, simplex and beta rectangular distributions and, in their respective augmented zero and/or one versions. We consider predictors, with a parametric and a nonparametric part, for the mean, and parametric predictors for the dispersion / precision and probabilities of occurrences of zeros and ones (for the augmented distributions). One of the advantages of this structure is to model linear and non-linear phenomena, properly. In the second approach, we focus on dependent data. In a similar way to the previous case, regression models were developed based on the aforementioned distributions, also under the semi-parametric approach for the mean (a parametric part and a nonparametric part for the predictor) and parametric for the other quantities, considering correlation structures via Generalized Estimation Equations (GEE). The two approaches were developed considering symmetric and asymmetric link functions, which correspond to generalizations of existing methodologies. In addition to the new model classes, structural selection, model fit assessment tools, measures of global and local influence, were developed for all models, for both approaches. In addition, appropriate simulation studies and real data analysis were conducted to illustrate the potential of the developed methodologies (AU) | |
| FAPESP's process: | 18/26784-1 - Parametric and semi-parametric modelos for longitudinal limited response |
| Grantee: | Áurea Fonseca Lopes Galindo |
| Support Opportunities: | Scholarships in Brazil - Master |