Λέσχη Φίλων Στατιστικής - GrStats forum
(20100527)AUEB SEMINARS - Nonlinear regression with missing responses, Athens - Greece Forumgrstats

Join the forum, it's quick and easy

Λέσχη Φίλων Στατιστικής - GrStats forum
(20100527)AUEB SEMINARS - Nonlinear regression with missing responses, Athens - Greece Forumgrstats
Λέσχη Φίλων Στατιστικής - GrStats forum
Would you like to react to this message? Create an account in a few clicks or log in to continue.
Για προβλήματα εγγραφής και άλλες πληροφορίες επικοινωνήστε με : grstats.forum@gmail.com ή grstats@stat-athens.aueb.gr

Go down
grstats
grstats
Posts : 967
Join date : 2009-10-21
http://stat-athens.aueb.gr/~grstats/

(20100527)AUEB SEMINARS - Nonlinear regression with missing responses, Athens - Greece Empty (20100527)AUEB SEMINARS - Nonlinear regression with missing responses, Athens - Greece

Wed 19 May 2010 - 13:19
Σεμινάριο Πεμπτη 27/05/2010 ωρα 15.00

Την Πέµπτη 27/05/2010, και ώρα 15.00-16.00 θα γίνει σεµινάριο από την Ursula U.
Muller (Department of Statistics, Texas A&M University) µε τίτλο
«Nonlinear regression with missing responses».
Το σεμινάριο θα δοθεί στην αίθουσα του 5ου ορόφου στην πτέρυγα Αντωνιάδου.
Ακλουθεί abstract

Nonlinear regression with missing responses

Uschi Muller-Harknett (Ursula U. Muller)
Department of Statistics, Texas A & M University
College Station, TX 77843-3143, USA
uschi@stat.tamu.edu
http://www.stat.tamu.edu/~uschi/

My talk will focus on linear and nonlinear regression, with a response
variable that is allowed to be ``missing at random''. My only structural
assumptions on the distribution of the variables are that the errors have
mean zero and are independent of the covariates. The independence
assumption is important: it enables us to construct easy-to-implement
estimators for expectations of the joint distribution that only require
a root-n consistent estimator for the parameter vector, but no elaborate
estimation techniques for the nonparametric parts. The independence assumption
further allows us to construct estimators for the response density that converge
at a faster rate than the usual local smoothing methods. The idea is to write the
density as an integral of the error distribution which can be estimated by
plugging in residual-based kernel estimators. For an appropriate class of regression
functions, the proposed density estimators are consistent and converge with the optimal
parametric rate $n^$. Moreover, both the estimator of the joint distribution
and the density estimator are proved to be efficient (in the sense of Hajek
and Le Cam) if an efficient estimator for the regression parameter is used.
The construction of such an estimator will also be addressed.
Back to top
Permissions in this forum:
You cannot reply to topics in this forum