[Forum SIS] avviso di seminario

[Brunero Liseo]

******************************
*     AVVISO DI SEMINARIO          *
******************************



La prof. Rebecca C. Steorts, del Deparment of Statistics, Carnegie
Mellon University, Pittsburgh, USA

terrà un seminario dal titolo:
Constrained Smoothed Bayesian Estimation

Segue abstract in calce


QUANDO: 16 maggio 2014 ore 14.30-15.30
DOVE:  Aula FANFANI, Dipartimento MEMOTEF,
facoltà di Economia, Sapienza.
Viale del Castro Laurenziano 9, quinto piano.


Tutti gli interessati sono invitati a partecipare

Brunero LISEO





ABSTRACT
 We develop methods of constrained Bayes estimation for small-area
estimation. We deal with two kinds of constraints: those that require
smoothness with respect to some form of similarity across areas, such as
geographic proximity or a clustering by covariates; and bench-marking
constraints, requiring (weighted) means and variances of estimates to agree
across levels of aggregation, or with external sources of information. We
develop our tools for constrained estimation both geometrically, by
projecting the
unconstrained Bayes estimate into the feasible set, and decision-theoretically,
by minimizing the posterior risk. We show that our
constrained estimators can be obtained as solutions to tractable
convex optimization
problems, and in some cases get closed-form solutions. Mean-squared errors
of the constrained estimators are calculated via bootstrapping. Our
techniques are free of distributional assumptions, and equally applicable
whether the estimator is linear or non-linear, univariate or multivariate.
We illustrated our methodology by applying it to data from the U.S.
Census's Small Area Income and Poverty Estimates program.


-- 

============================================
Brunero Liseo
*Dip. di metodi e modelli per il territorio, l'economia e la finanza *
*Sapienza Università di Roma*

*Viale Castro Laurenziano, 9 Roma I-00161 *Tel. +39 06 49766973
Fax +39 06 4957606
http://geostasto.eco.uniroma1.it/utenti/liseo
=============================================
-------------- parte successiva --------------
Un allegato HTML è stato rimosso...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20140512/12df11fc/attachment-0001.html>