Information détaillée concernant le cours
École d’été 2018
02-05 Septembre 2018
M. Yves Tillé, UNINE (Président); Mme Caroline Gillardin, UNINE (coordinatrice)
Prof. Yoav Benjamini, Tel Aviv University (invité par l'EPFL); Prof. Claudia Czado, Technische Universität München; Prof. Ingrid Van Keilegom, Katholieke Universiteit Leuven
Professor Yoav Benjamini
(Tel Aviv University)
Title : Selective inference
1.The problem of selective inference when facing multiplicity
We shall review the basic problem, its relationship to the replicability problems in Science, and the more traditional approach of simultaneous inference.
2.An in-depth tour of the False Discovery Rate (FDR) and False Coverage Rate (FCR).
We shall get into the concepts and particular methods, including adaptive methods, which try to assure that inferential properties hold on the average over the selected. We shall discuss the connections between dependency assumptions and the validity of the methods, as well as resampling methods.
3.Some recent advances in addressing the problem of selective inference and open problems.
We shall discuss topics such as testing hierarchical systems of hypotheses, the knockoff approach and the conditional approach for addressing selective inference.
Professor Claudia Czado
(Technische Universität München)
Title : Analyzing dependent data with vine copulas
This course is designed for graduate students and researchers who are interested in usingcopula based models for multivariate data structures. It provides a step to step introduction to the class of vine copulas and their statistical inference. This class of flexible copula models has become very popular in the last years for many applications in diverse fields such as finance, insurance, hydrology, marketing, engineering, chemistry, aviation, climatology and health.The popularity of vines copulas is due to the fact that it allows in addition to the separation of margins and dependence by the copula approach, tail asymmetries and separate multivariate component modeling. This is accommodated by constructing multivariate copulas using only bivariate building blocks which can be selected independently. These building blocks are glued together to valid multivariate copulas by appropriate condition-
Professor Ingrid Van Keilegom
(Katholieke Universiteit Leuven)
Title : Survival analysis : from basic concepts to open research questions
In the first part of the course some basic concepts of survival analysis will be reviewed, like the concepts of right censoring and left truncation, some common parametric distribution functions in survival analysis, nonparametric estimation of basic quantities (Kaplan-Meier estimator of the survival distribution, Nelson-Aalen estimator of the cumulative hazard function,…), hypothesis testing regarding the equality of two or more survival curves, proportional hazards models, accelerated failure time models, etc.
The second part of the course will treat a number of more specific topics in survival analysis, that are in full development :
(1) Cure models : these are survival models used in situations where a certain proportion of the subjects under study are not susceptible to the event of interest (i.e. they have an infinite survival time). These models are of interest e.g. in cancer studies where one is interested in the time until recurrence of the cancer. Those individuals who are cured of their cancer will never have a relapse and will hence have an infinite event time. We will review some of the existing models and estimation methods in this context.
(2) Dependent censoring : Most models and methods in survival analysis assume that the survival time and the censoring time are independent random variables. This assumption is made for identifiability reasons, but is not satisfied in a number of practical situations. We will discuss how existing models can be adapted to take dependent censoring into account.
(3) Measurement errors : In survival analysis, as in many other areas of statistics, it often happens that explanatory variables in a regression model are measured with some error (e.g. blood pressure, weight, wage,…). Taking measurement errors into account is essential to do valid estimation and inference. We will review some of the existing models and methods for common regression models in survival analysis.
While the second part focusses on existing literature in the three stated sub-areas of survival analysis, the third part of the course will go one step further and will handle some of the open issues and research questions in these three areas. Since these areas are in full development, many aspects are still unexplored and worth investigating.
Eurotel Victoria Villars
Doctorant CUSO chambre double: 200 CHF
Versement sur compte postal: