October 23rd 2012
Faculty of Economic and Business Sciences | Salón de Grados
Conditional Estimation of the Bivariate Distribution Under Dependent Right Censoring
2012/10/23 – 12:00 h | Ana Moreira, University of Minho (Portugal)
Abstract
In many medical studies individuals can experience several events across a follow-up study. In these studies, the times between two consecutive events are often of interest and lead to problems that have received much attention. Most of the times, one will be interested in describing the distribution of the joint gap times, the marginal distribution of the gap times but also the correlation structure among them. In recent years significant contributions have been made regarding this topic. However, most approaches assume independent censoring and do not account for the influence of covariates. This talk introduces two estimators that account for dependent censoring while including covariate information. A real data illustration is included. Joint work with Luís Machado.
Advances in survival data: parametric versus nonparametric truncation
2012/10/23 – 12:45 h | Jacobo de Uña-Álvarez, University of Vigo
Abstract
In Survival Analysis and other applied fields, the observation of the lifetime or event time of ultimate interest is influenced by (besides of censoring) left-truncation. The most important example of this situation is given by cross-sectional sampling, that is, when only individuals in progress at the sampling date (prevalent cases) enter the data basis. Left-truncation induces an observational bias in the lifetime, so the application of the standard Kaplan-Meier method overestimates the survival probability. To overcome this issue, several estimators have been proposed in the last decades. In particular, it has been shown that information on the distribution of the truncation time may be used to construct estimators with a smaller variance. Special attention has been recently paid to estimation with left-truncated and right-censored data when the truncation time is uniformly distributed, leading to the so-called length-biased data. In this talk I will review some of these approaches, with a remark on their relative advantages and disadvantages. As an outstanding open problem, I will discuss the construction of a semiparametric estimator which may outperform the nonparametric maximum likelihood estimator when the researcher has partial information on the truncation distribution.