[Forum SIS] Corso FREQUENTIST LIMITS FROM BAYESIAN STATISTICS

Matteo Ruggiero matteo.ruggiero a unito.it
Ven 19 Nov 2021 11:13:20 CET


Cari colleghi,

nell’ambito del programma di Visiting Professors della Laurea Magistrale in Stochastics and Data Science dell’Università di Torino (alla pagina https://www.master-sds.unito.it/go/visiting <https://www.master-sds.unito.it/go/visiting> il programma completo per il presente a.a.), con piacere annunciamo il seguente corso:

--------------------------------------

Bas KLEIJN (University of Amsterdam)

FREQUENTIST LIMITS FROM BAYESIAN STATISTICS

We look at the role that the posterior can play in frequentist statistics, with special attention for large-sample limits. There exists an intimate relationship between hypothesis testing and concentration of posterior measure, which combines with a notion called remote contiguity to obtain frequentist convergence results from sequences of Bayesian posterior measures. We shall cover frequentist forms of posterior consistent estimation, rates of posterior convergence, as well as asymptotic uncertainty quantification by enlargement of credible sets and hypothesis testing/model selection with Bayes factors. 

Plan of the course: eight lectures of 45 minutes each:
Lec I - Bayesian Basics (Frequentist/Bayesian formalisms, estimation, coverage, testing)
Lec II - The Bernstein-von Mises theorem (Limit shape in smooth parametric models, semi-parametrics)
Lec III - Bayes and the Infinite (Consistency, Doob’s theorem, Schwartz’s theorem)
Lec IV - Posterior contraction (Barron, Walker, Ghosh-Ghosal-van der Vaart theorems)
Lec V - Tests and posteriors (Testing and posterior concentration, Doob’s theorem)
Lec VI - Frequentist validity of Bayesian limits (Remote contiguity and frequentist limits)
Lec VII - Posterior uncertainty quantification (How confidence sets arise from credible sets)
Lec VIII - Uniform and pointwise tests (Which hypotheses are asymptotically testable and which are not?)

Time permitting, we shall also consider

Lec IX - Bayesian tests and posterior model selection (Model selection with posteriors, some examples)

--------------------------------------

L'iniziativa è rivolta agli studenti del secondo anno della Laurea Magistrale ma la partecipazione è aperta a tutti gli interessati. 
Il corso si svolgerà in presenza nell’aula 30, terzo piano, C.so Unione Sovietica 218/bis, 10134, Torino (qui
https://www.unito.it/ateneo/gli-speciali/coronavirus-aggiornamenti-la-comunita-universitaria/rientri-accesso-alle <https://www.unito.it/ateneo/gli-speciali/coronavirus-aggiornamenti-la-comunita-universitaria/rientri-accesso-alle>
i requisiti per l’accesso come ingressi occasionali) secondo il seguente calendario:

- mar 30 Novembre, h.14-16
- mer 1 Dicembre, h.14-16
- mar 7 Dicembre, h.14-16
- gio 9 Dicembre, h.11.15-13

e sarà trasmesso in diretta streaming (senza registrazione) tramite la seguente riunione Webex, valida per tutte le lezioni:

https://unito.webex.com/unito/j.php?MTID=m508322b370362bfd0618d768e4772ac0 <https://unito.webex.com/unito/j.php?MTID=m508322b370362bfd0618d768e4772ac0>
Numero riunione: 2733 233 8719
Password: 9ahN5DGFRU3


Cordiali saluti,
Matteo Ruggiero

---
Matteo Ruggiero
University of Torino and Collegio Carlo Alberto
www.matteoruggiero.it <http://www.matteoruggiero.it/>
-------------- parte successiva --------------
Un allegato HTML è stato rimosso...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20211119/565fb998/attachment.html>


Maggiori informazioni sulla lista Sis