Κατεύθυνση Στατιστική και Επιχειρησιακή Έρευνα
Library of the School of Science

2012-06-28

2012

Λαμπρινάκου Φιόρη

Λουκία Μελιγκοτσίδου Λέκτορας ΕΚΠΑ

Μέθοδοι Markov Chain Monte Carlo σε Μοντέλα Ποσοστημοριακής Παλινδρόμησης

Greek

Regression analysis is the most prevalent technique in studying the mean
behavior of a dependent variable, given other independent variables which
affect it. However, it is often interesting to investigate this correlation
throughout the whole range of the dependent variable's distribution rather than
just on its mean. The answer to this question is given by quantile regression.
Bayesian approach to quantile regression is another current area that has
attracted intense interest, as it appears to have some advantages over
classical inference.One of these is that it provides estimation and forecasts,
which fully take into account parameter uncertainty. Finally a problem that we
often face is the selection of appropriate independent variables that will be
used in the model. When we have data from several independent variables it is
not necessary that all of them will be used in the final model. A quite common
method in selecting appropriate variables is the Stochastic Search Variable
Selection (SSVS).

The aim of this master thesis is (a) the presentation of classical quantile
regression (b) the presentation of Bayesian analysis in quantile regression
models and (c) the selection of variables using the SSVS method in quantile
regression models

Quantile regression, Bayesian inference, Markov Chain Monte Carlo, Stochastic Search Variable Selection

No

0

Yes

19

100

File access is restricted.

https://pergamos.lib.uoa.gr/uoa/dl/object/1318974