Banca de DEFESA: ERIKA RAYANNE FERNANDES DA SILVA

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : ERIKA RAYANNE FERNANDES DA SILVA
DATE: 18/02/2020
TIME: 09:00
LOCAL: Auditório do CCET
TITLE:

Beta modal regression model 

KEY WORDS:

Mode; Beta regression; Parametric modal regression.

PAGES: 65
BIG AREA: Ciências Exatas e da Terra
AREA: Probabilidade e Estatística
SUMMARY:

The beta regression model is a class of models used for continuous response variables
restricted to the interval (0,1), such as rates and proportions. Ferrari and Cribari-Neto
(2004) proposed a beta regression model that incorporates covariates in the mean of the
distribution using a generic link function. However, for studies which response variable
has asymmetry and / or discrepant values, this model may not be the most appropriated.
A more appropriated measure of central tendency in this kind of situation is the mode
of distribution because of its robustness to outliers and its easy interpretation in cases of
asymmetry. Zhou and Huang (2019) proposed a parametrization for the beta distribution
in terms of the mode and a precision parameter. Assuming that the response variable
follows this distribution, Zhou and Huang (2019) proposed a regression model for continuous
data in the interval (0,1), in order to be more robust to outliers. In this work,
we present a more complete study of this model properties and performance as well as
a comparison between this model and the model proposed by Ferrari and Cribari-Neto
(2004). We developed simulation studies to evaluate the maximum likelihood estimates in
cases of asymmetry and the sensitivity to outliers when come perturbation patterns are
imposed. Furthermore, we proposed and evaluated three residuals to this class of models.
Our simulation studies suggest that the model developed to the mode has a good performance
on symmetrical and asymmetrical data and in most scenarios it performs better
in the presence of outliers than the beta regression model that considers the mean. In
addition, we present two applications to real dataset and a t comparison of the models
that consider the mean and the mode.

BANKING MEMBERS:
Interna - 734492 - DIONE MARIA VALENCA
Interno - 2612836 - FRANCISCO MOISES CANDIDO DE MEDEIROS
Externo à Instituição - JEREMIAS DA SILVA LEÃO - UFAM
Presidente - 1023112 - MARCELO BOURGUIGNON PEREIRA