Advances and Applications in Statistics
Volume 50, Issue 5, Pages 367 - 384
(May 2017) http://dx.doi.org/10.17654/AS050050367 |
|
GENERALIZED WEIBULL LINEAR MODELS WITH DIFFERENT LINK FUNCTIONS
Afsana-Al-Sharmin and M. Ataharul Islam
|
Abstract: The Weibull distribution is one of the most extensively used distributions in survival analysis. Use of the Weibull distribution can be improved substantially if regression models analogous to the class of generalized linear models (GLMs) are employed. In this paper, regression models for the Weibull distributions are examined for different types of link functions and compared with a framework of reparameterized Weibull regression model. This framework is similar to the GLM. The reparameterized Weibull regression model using log link function is applied to the infant survival data and is compared with alternative modeling procedures. It is found that the reparameterized model performs better which is shown by different goodness of fit measures such as deviance, AIC and BIC. A simulation study is also conducted and it is revealed that the generalized Weibull regression model with reparameterized log link function performs better than the alternative models with different link functions. |
Keywords and phrases: Weibull distribution, generalized Weibull regression model, negative link, log link, identity link, reparameterized. |
|
Number of Downloads: 479 | Number of Views: 1583 |
|