A completamento dell'annuncio sul ciclo di seminari che il prof. James O. Berger terrà presso la facoltà di Economia Sapienza Università di Roma secondo il calendario seguente
Bayesian Model Selection
Aula 6B, piano terra
F acoltà di Economia, Via del Castro Laurenziano, 9,
Lunedì 22 gennaio ore 11-13
Martedì 23 gennaio ore 16.30-18
Mercoledì 24 gennaio 11.30-13
Mercoledì 24 gennaio 14.30-16
si allega abstract dei singoli seminari
Abstract: These lectures address the Bayesian approach to hypothesis testing and model uncertainty, with extensive comparisons to the classical approaches to these subjects.
The first lecture, "Introduction to Bayesian Hypothesis Testing," begins with a brief introduction to Bayesian analysis, for students who have not been exposed to the subject. The lecture then goes on to highlight the main issues, through a discussion of p-values (the classical approach to hypothesis testing) versus Bayes factors (the Bayesian approach), primarily done through pedagogical examples. The first lecture (after a break) will go on to present the formalism of Bayesian hypothesis testing, with examples.
The second lecture "Interfaces Between Testing Paradigms," will take a step back to the historical development of statistics, and describe how the classical and Bayesian approaches of hypothesis testing developed, and show that there is currently methodology that is completely compatible with both perspectives.
The third lecture "Essentials of Bayesian Model Uncertainty," extends the first two lectures to the subject of dealing with many possible statistical models. Interestingly, none of the basic methodology changes - it just becomes more difficult to implement.
The final lecture focuses on "Variable Selection in the Linear Model," arguably the most important statistical problem, and discusses Bayesian resolutions that have amazing properties, as well as software that implements the methodology.