Vol.:(0123456789)
Journal of Statistical Theory and Practice (2019) 13:27
https://doi.org/10.1007/s42519-018-0026-3
1 3
ORIGINAL ARTICLE
Log‑Burr XII Gamma–Weibull Regression Model
with Random Efects and Censored Data
Elizabeth M. Hashimoto
1
· Giovana O. Silva
2
· Edwin M. M. Ortega
3
·
Gauss M. Cordeiro
4
© Grace Scientific Publishing 2018
Abstract
It may happen in some applications that the assumption of independence of sur-
vival times does not hold. Thus, we propose a new log-Burr XII regression model
with log-gamma–Weibull distributions for the random efects. The maximum like-
lihood method is used to estimate the model parameters based on the Gauss–Her-
mite numerical integration technique. For diferent parameter settings, sample sizes,
censoring percentages and correlated data, various simulations are performed. Some
global-infuence measurements are also investigated. In order to assess the robust-
ness of the maximum likelihood estimators, we evaluate local infuence on the
estimates of the parameters under diferent perturbation schemes. We illustrate the
importance of the new model by means of a real data set in analysis of experiments.
Keywords Censored data · Log-gamma–Weibull distribution · Random efect ·
Regression model
1 Introduction
In regression models for survival analysis, the usual assumption that the survival
times of distinct elements are independent may not be valid in some applications.
Sometimes the elements observed form a group, and thus, the survival times within
each group might not be mutually independent. This is the case, for example, of
data observed about people of the same family or animals of the same litter. Besides
* Edwin M. M. Ortega
edwin@usp.br
1
Departamento Acadêmico de Matemática, UTFPR, Curitiba, Brazil
2
Departamento de Estatística, UFBA, Salvador, Brazil
3
Departamento de Ciências Exatas, ESALQ - USP, Av. Pádua Dias 11, Caixa Postal 9,
13418-900 Piracicaba, São Paulo, Brazil
4
Departamento de Estatística, UFPE, Recife, Brazil