MEIO - Summer School - Bayesian Data Analysis

Títol del curs	Bayesian Data Analysis An Introduction to Bayesian Statistics with Applications in Biostatistics and Epidemiology
Impartit per	Emmanuel Lesaffre
Llengua del curs	Anglès
Dates i horaris del curs	Dia 17 de setembre: 10-12h & 13-15h Dies 18, 19, 20 & 25 de setembre: 9h-11h & 12-14h
Tipus d'activitat i càrrega lectiva	Curs de 20 hores
Reconeixement acadèmic	2,5 ECTS com a assignatures optatives per als estudiants del MEIO, 2,5 crèdits com a ALE per als de l'LCTE, i 2 crèdits per als de Doctorat
Destinataris	Estudiants del MEIO, de l'LCTE i estudiants de Doctorat
Data de matrícula	El 15 de maig de 2007 - de 11 a 13h i de 16 a 18h
Presentació	Aquest curs de 2 setmanes de durada compta amb la participació del Prof. Lesaffre, catedràtic de la Universitat Catòlica de Leuven a Bèlgica i un gran científic en l’àrea de l’estadística i en particular en el desenvolupament de metodologies basades, entre d’altres coses, en la filosofia bayesiana. El curs Bayesian Data Analysis completa el curs d’Inferència Bayesiana que s’imparteix en el segon semestre del Màster. L’enfocament que el Prof. Lesaffre dóna al seu curs és primordialment pràctic, motivat en estudis de casos i basat en la seva gran experiència aplicada.
Objectius del curs	Presentar la filosofia i la terminologia Bayesiana, tot contrastant-la amb la freqüentista. Repassar les tècniques numèriques de càlcul Bayesià més importants. Mostrar el programari que es fa servir per ajustar models Bayesians.
Continguts	Bayesian statistics has received a great deal of attention as a method to tackle complex statistical analyses. The purpose of this course is to gradually and smoothly introduce the participants into the Bayesian philosophy and terminology. The early examples are simple but inspired by clinical trial examples and simple epidemiological studies. To this end the menu-driven software FirstBayes will be employed. While in general the technical level of the course is low, the different numerical techniques to calculate (or to sample from) the posterior distribution will be treated. The recent Markov Chain Monte Carlo (MCMC) techniques will be explored extensively thereby trying to keep the technical level low. Knowledge of R or S+ is useful in this part of the course in order to exercise in writing some simple MCMC programs. Further, WINBUGS will be used to illustrate the power of the MCMC methods with a variety of biostatistical and epidemiological applications. For instance we will consider the use of Bayesian methods in disease mapping, meta-analyses, survival analysis, errors-in-variables problems, missing data problems, clinical trials such as the use of Bayesian methods for the planning of trials, in compliance problems, data and safety monitoring boards, etc.
Planificació	Day 1: Frequentist and likelihood approach in contrast to Bayesian philosophy Bayes' theorem and Bayesian summary statistics: binomial, normal and Poisson likelihood Day 2 Tutorial with FirstBayes Sampling from the posterior distribution with applications in R (or S+) Towards real life problems: multi-parameter models Day 3: Choosing the prior distribution Markov Chain Monte Carlo methods: Gibbs and Metropolis sampling Day 4: First examples with R and WINBUGS: Bayesian meta-analyses, disease mapping, the use of historical data in an epidemiological study Day 5: More advanced analyses with WINBUGS: o correction of scoring errors in a dental epidemiological study o repeated measurements studies o the estimation of prevalence using probabilistic constraints o survival analyses o etc.