INTERNATIONAL SEMINAR ON MULTI-STATE MODELS
Vigo, March 26 2014
Salón de Grados de la Facultad de Ciencias Económicas y
Universidad de Vigo
Campus Universitario Lagoas-Marcosende, 36310 Vigo – Spain
Jacobo de Uña Álvarez
Grupo de Investigación SiDOR & Departamento de Estadística e Investigación Operativa
Universidad de Vigo
Funded by: Programa de Consolidación y Estructuración de Unidades de Investigación Competitivas del SUG, Xunta de Galicia
10:30 – 11:15 “The Jackknife estimate of variance for transition probabilities in the non-Markov illness-death model”, Leyla Azarang (University of Vigo, Spain)
Abstract: Non-parametric estimators of transition probabilities for non-Markov illness-death model were introduced by Meira-Machado et al. (2006). In this seminar, to compute confidence intervals for these estimators, the problem of consistently estimating the limit variance will be explored. To estimate the variance, the limit variances and covariance for some special bivariate Kaplan-Meier integrals with covariates, where the marginal distribution of the total survival time of the process weights the data and the sojourn time in the initial state plays the role of covariate, are approximated using the Jackknife method. Through some simulation study, the new asymptotic result will be discussed.
11:15 – 11:45 Coffee-break
11:45 – 12:30 “A review of the estimation of transition probabilities in non-Markov models and related quantities”, Luís Filipe Meira-Machado (University of Minho, Portugal) http://w3.math.uminho.pt/~lmachado/
Abstract: One important goal in multi-state modeling is the estimation of transition probabilities. In longitudinal medical studies these quantities are particularly of interest since they allow for long-term predictions of the process. In recent years significant contributions have been made regarding this topic. First, new estimators have been introduced for non-Markov processes; Second, recent methods were proposed that deal with the problem of dependent censoring and account for the influence of covariates. Other important targets include the state occupation probabilities (which can be seen as a particular case of the transition probabilities), the cumulative incidence function and the waiting time distributions. In this talk new estimators for (some of) these quantities will be introduced. Simulations demonstrate that the new estimators may outperform the later estimators. An illustration through real data analysis is included.
12:30 – 13:15 "Robust estimates of state occupancy and transition probabilities for Non-Markov multi-state models", Andrew Titman (University of Lancaster, UK) http://www.maths.lancs.ac.uk/~titman/
The seminar will consider approaches for consistently estimating the marginal
state occupancy probabilities and transition probabilities in multi-state
models, considering both continuously observed and interval censored data. In
each case, the objective is to obtain estimates that remain consistent for
non-Markov models. For continuously observed data up to right-censoring
an extension of the work of Meira-Machado et al
(2006) and Allignol et al (2013), to allow estimation
of marginal transition probabilities for general non-Markov models will be
presented. The proposed estimator involves the cumulative incidence functions
of a constructed competing risks process.
In the latter part of the talk, a computationally simple method for obtaining non-parametric estimates of the state occupation probabilities for progressive multi-state models under interval censoring will be presented. The method involves estimating time-to-absorption using standard survival techniques and using methods for current status data to estimate the conditional probabilities of occupancy in non-absorbing states. The basic estimator is valid under the assumption that the observation times are completely independent of the multi-state process. An estimator valid under general non-informative observation can be constructed through the use of inverse intensity weighting. Some discussion of the difficulties of estimating transition probabilities, as opposed to occupation probabilities, in the interval-censored case will be also be given.
There is no registration fee. However, due to organizational issues, sending an e-mail to firstname.lastname@example.org with subject: Registration MSM2014 before Wednesday March 19 at 14:00 is mandatory. Please indicate full name and affiliation (students at the Galician interuniversity Statistics Master Program or at the Galician interuniversity Statistics and OR Doctorate Program are particularly welcome).
La inscripción es gratuita. Por motivos de organización, es necesario para inscribirse enviar antes del miércoles 19 de marzo a las 14:00 horas un e-mail a email@example.com con tema: Inscripción MSM2014 indicando nombre completo y afiliación (los estudiantes del Máster interuniversitario en Técnicas Estadísticas y del Programa de Doctorado interuniversitario en Estadística e Investigación Operativa son especialmente bienvenidos).