DivulgaÃÃo de seminÃrio

[antonio eduardo gomes <aegomes@unb.br>]

PROGRAMA DE PÃS-GRADUAÃÃO EM ESTATÃSTICA
        UNIVERSIDADE DE BRASÃLIA

Data: 01/02/2011
horario: 16:30
Local: Dep. de EstatÃstica - sala 34

Palestrante: Rodrigo Tsai
             Doutorando em EstatÃstica - Unicamp

TÃtulo: Polyhazard Models with Dependent Causes

Resumo:
Polyhazard models are a flexible family for fitting lifetime data.
The main advantages over the single hazard models is the flexibility
to represent hazard rates functions with unusual shapes and the
facility to include covariates. The main goal of this paper is to include
dependence among the latent causes of failure. The dependence is modeled
trough copula functions. The choice of the copula function and the
latent hazard functions result in a quite flexible class of survival functions
which is able to represent hazard rates functions with unusual shapes as
bathtub or multimodal curves and also containing local effects given by
the competition between the risks. The model identification and estimation
are discussed and the model is applied to two set of simulated data
and to the unemployment duration of German socially insured workforce.

* Trabalho conjunto com o Prof. Luiz Koodi Hotta (Dep. de EstatÃstica  
- Unicamp)