東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Model combining and its applications...

Liu, Song.

Linked to FindBook

Google Book

Amazon

博客來

Model combining and its applications: Longitudinal and semiparametric models.

Record Type:	Language materials, printed : Monograph/item
Title/Author:	Model combining and its applications: Longitudinal and semiparametric models./
Author:	Liu, Song.
Description:	74 p.
Notes:	Adviser: Yuhong Yang.
Contained By:	Dissertation Abstracts International67-11B.
Subject:	Statistics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3243354
ISBN:	9780542992261

Model combining and its applications: Longitudinal and semiparametric models.
Liu, Song.

Model combining and its applications: Longitudinal and semiparametric models. - 74 p.

Adviser: Yuhong Yang.

Thesis (Ph.D.)--University of Minnesota, 2006.

Based on our results, we recommend that model selection diagnostics be done when model selection is involved and model combining should be applied when model selection is too uncertain for the task of interest.

ISBN: 9780542992261Subjects--Topical Terms:

517247
Statistics.

Model combining and its applications: Longitudinal and semiparametric models.
LDR:03063nam 2200301 a 45 001 963921
005 20110831
008 110831s2006 ||||||||||||||||| ||eng d
020 $a 9780542992261
035 $a (UMI)AAI3243354
035 $a AAI3243354
040 $a UMI $c UMI
100 1 $a Liu, Song. $3 1275819
245 1 0 $a Model combining and its applications: Longitudinal and semiparametric models.
300 $a 74 p.
500 $a Adviser: Yuhong Yang.
500 $a Source: Dissertation Abstracts International, Volume: 67-11, Section: B, page: 6486.
502 $a Thesis (Ph.D.)--University of Minnesota, 2006.
520 $a Based on our results, we recommend that model selection diagnostics be done when model selection is involved and model combining should be applied when model selection is too uncertain for the task of interest.
520 $a Model selection delivers both simple interpretability and accurate information about the randomness of the data if it is reliable. However; when there is large amount of uncertainty involved in model selection, insisting on simple interpretability is not appropriate and conclusions based on the selected model are usually overly optimistic or misleading. In such a situation, a more reliable approach such as model averaging/combining should be considered.
520 $a Although model selection uncertainty has now been well recognized, proper diagnostics of model selection, a key for choosing between model selection and model combining, have not been seriously studied. In this work; we propose measures that, are sensible for variable selection or regression estimation that involves model selection. In particular; a variable selection standard deviation is defined in terms of a distance between the candidate models and a sensible weighting distribution on them. Compared to bootstrap instability measures; it overcomes the over-aggressiveness in declaring a severe selection uncertainty. On the other hand, when variable selection standard deviation is high; we cannot trust the selected variables as the most important ones.
520 $a We propose new methods for combining models. One is suitable for longitudinal data where observations within the same subject are correlated; which requires one to take care of the covariance structure when combining models. The other is appropriate for semiparametric models where the parametric part of the model is of much interest. We investigate the properties of our combining strategy both theoretically and empirically. The theorems guarantee that the combined estimator achieves the optimal rate of convergence without knowing which model works the best. The empirical studies confirm that the combined estimator has a better prediction performance than the single selected model when model selection uncertainty is high.
590 $a School code: 0130.
650 4 $a Statistics. $3 517247
690 $a 0463
710 2 $a University of Minnesota. $3 676231
773 0 $t Dissertation Abstracts International $g 67-11B.
790 $a 0130
790 1 0 $a Yang, Yuhong, $e advisor
791 $a Ph.D.
792 $a 2006
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3243354