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

Towards complete results for some in...

Meng, Xiao-Li.

Linked to FindBook

Google Book

Amazon

博客來

Towards complete results for some incomplete-data problems.

Record Type:	Language materials, printed : Monograph/item
Title/Author:	Towards complete results for some incomplete-data problems./
Author:	Meng, Xiao-Li.
Description:	145 p.
Notes:	Source: Dissertation Abstracts International, Volume: 51-12, Section: B, page: 5960.
Contained By:	Dissertation Abstracts International51-12B.
Subject:	Biology, Biostatistics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=9113196

Towards complete results for some incomplete-data problems.
Meng, Xiao-Li.

Towards complete results for some incomplete-data problems. - 145 p.

Source: Dissertation Abstracts International, Volume: 51-12, Section: B, page: 5960.

Thesis (Ph.D.)--Harvard University, 1990.

Three different topics in incomplete-data problems are explored in this thesis. The first topic is the EM algorithm, a very general and often remarkably simple iterative algorithm for computing maximum likelihood estimates and posterior modes in incomplete-data problems. Some general results on the global rate and component-wise rates of convergence of EM are presented. The rate of convergence of EM is then used to develop a supplemented EM algorithm (SEM) for computing the observed-data observed information matrix or its inverse. Finally, a promising generalized EM algorithm, the ECM algorithm, is proposed and investigated. This new algorithm preserves the nice convergence properties of EM, but extends the power of EM since its maximization step is more flexible.Subjects--Topical Terms:

1018416
Biology, Biostatistics.

Towards complete results for some incomplete-data problems.
LDR:02937nam 2200301 4500 001 1394008
005 20110419112656.5
008 130515s1990 ||||||||||||||||| ||eng d
035 $a (UMI)AAI9113196
035 $a AAI9113196
040 $a UMI $c UMI
100 1 $a Meng, Xiao-Li. $3 753103
245 1 0 $a Towards complete results for some incomplete-data problems.
300 $a 145 p.
500 $a Source: Dissertation Abstracts International, Volume: 51-12, Section: B, page: 5960.
500 $a Adviser: Donald B. Rubin.
502 $a Thesis (Ph.D.)--Harvard University, 1990.
520 $a Three different topics in incomplete-data problems are explored in this thesis. The first topic is the EM algorithm, a very general and often remarkably simple iterative algorithm for computing maximum likelihood estimates and posterior modes in incomplete-data problems. Some general results on the global rate and component-wise rates of convergence of EM are presented. The rate of convergence of EM is then used to develop a supplemented EM algorithm (SEM) for computing the observed-data observed information matrix or its inverse. Finally, a promising generalized EM algorithm, the ECM algorithm, is proposed and investigated. This new algorithm preserves the nice convergence properties of EM, but extends the power of EM since its maximization step is more flexible.
520 $a The second topic is multiple imputation, another very general and computationally efficient technique for handling incomplete-data or missing-data problems. Various procedures are studied for obtaining significance levels from the multiply-imputed data sets. A modification of the current best moment-based procedure is presented along with its theoretical justification. The current best chi-squared-based procedure is further justified when the conditions under which the procedure was originally derived are violated. A simple procedure is introduced and illustrated for performing likelihood ratio tests with multiple imputation. This new procedure is asymptotically equivalent to the current best procedure, but is computationally more efficient.
520 $a The third topic deals with a specific form of incomplete data, censored data. A generalized Hodges-Lehmann estimator for estimating the treatment effect in the two-sample problem with right censoring is studied. Its strong consistency is proved under the minimal possible conditions, and its asymptotic normality is established regardless of the censoring mechanisms. As a byproduct, Bahadur's result on the oscillation behavior of the empirical process is extended to the Kaplan-Meier process under the same conditions.
590 $a School code: 0084.
650 4 $a Biology, Biostatistics. $3 1018416
650 4 $a Statistics. $3 517247
650 4 $a Computer Science. $3 626642
690 $a 0308
690 $a 0463
690 $a 0984
710 2 $a Harvard University. $3 528741
773 0 $t Dissertation Abstracts International $g 51-12B.
790 1 0 $a Rubin, Donald B., $e advisor
790 $a 0084
791 $a Ph.D.
792 $a 1990
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=9113196