東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Statistical assessment of difference...

Howell, Patrick.

FindBook

Google Book

Amazon

博客來

Statistical assessment of differences in information retrieval effectiveness.

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Statistical assessment of differences in information retrieval effectiveness./
作者:	Howell, Patrick.
面頁冊數:	200 p.
附註:	Adviser: Robert M. Losee.
Contained By:	Dissertation Abstracts International63-11A.
標題:	Information Science. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3070861
ISBN:	0493906622

Statistical assessment of differences in information retrieval effectiveness.
Howell, Patrick.

Statistical assessment of differences in information retrieval effectiveness. - 200 p.

Adviser: Robert M. Losee.

Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2002.

Statistical evaluation of differences in retrieval effectiveness has been considered problematic primarily because the use of most commonly known statistical tests requires certain assumptions be made that information retrieval data have not been demonstrated to satisfy. This report identifies a mathematical model of retrieval system behavior consistent with fundamental principles of probability theory and presents a conceptual framework within which multiple outcomes from document retrieval experimentation may be evaluated for statistical significance of difference in effectiveness. Within this framework, large-sample approximate, and exact conditional tests for the statistical significance of difference in effectiveness levels among multiple retrieval outcomes are derived. A Monte Carlo simulation study based on a sample-without-replacement urn model of document retrieval processes was conducted to establish standards with which to compare probabilities for differences in retrieval effectiveness as calculated by the exact conditional method. No compelling evidence was found to suggest that the underlying probabilistic model of the proposed exact test for statistical significance of difference in retrieval effectiveness differs substantially from that of the conceptual urn model of independent document retrieval systems as realized in the Monte Carlo simulations. When retrieval results can be presented as cell frequencies of multiple two-by-two tables, the suggested exact conditional test was shown to provide for the evaluation of statistical significance of differences in retrieval effectiveness; however, some results caused questions to be raised of the test's power. The proposed large-sample and exact statistical tests for differences in retrieval effectiveness are demonstrated on a subset of data from the second Cranfield project. Despite concerns about possible violations of assumptions, probabilities computed by two of the large-sample tests studied exhibited notable accordance with probabilities determined by the proposed exact conditional test.

ISBN: 0493906622Subjects--Topical Terms:

1017528
Information Science.

Statistical assessment of differences in information retrieval effectiveness.
LDR:03013nam 2200289 a 45 001 937825
005 20110511
008 110511s2002 eng d
020 $a 0493906622
035 $a (UnM)AAI3070861
035 $a AAI3070861
040 $a UnM $c UnM
100 1 $a Howell, Patrick. $3 1261681
245 1 0 $a Statistical assessment of differences in information retrieval effectiveness.
300 $a 200 p.
500 $a Adviser: Robert M. Losee.
500 $a Source: Dissertation Abstracts International, Volume: 63-11, Section: A, page: 3770.
502 $a Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2002.
520 $a Statistical evaluation of differences in retrieval effectiveness has been considered problematic primarily because the use of most commonly known statistical tests requires certain assumptions be made that information retrieval data have not been demonstrated to satisfy. This report identifies a mathematical model of retrieval system behavior consistent with fundamental principles of probability theory and presents a conceptual framework within which multiple outcomes from document retrieval experimentation may be evaluated for statistical significance of difference in effectiveness. Within this framework, large-sample approximate, and exact conditional tests for the statistical significance of difference in effectiveness levels among multiple retrieval outcomes are derived. A Monte Carlo simulation study based on a sample-without-replacement urn model of document retrieval processes was conducted to establish standards with which to compare probabilities for differences in retrieval effectiveness as calculated by the exact conditional method. No compelling evidence was found to suggest that the underlying probabilistic model of the proposed exact test for statistical significance of difference in retrieval effectiveness differs substantially from that of the conceptual urn model of independent document retrieval systems as realized in the Monte Carlo simulations. When retrieval results can be presented as cell frequencies of multiple two-by-two tables, the suggested exact conditional test was shown to provide for the evaluation of statistical significance of differences in retrieval effectiveness; however, some results caused questions to be raised of the test's power. The proposed large-sample and exact statistical tests for differences in retrieval effectiveness are demonstrated on a subset of data from the second Cranfield project. Despite concerns about possible violations of assumptions, probabilities computed by two of the large-sample tests studied exhibited notable accordance with probabilities determined by the proposed exact conditional test.
590 $a School code: 0153.
650 4 $a Information Science. $3 1017528
650 4 $a Library Science. $3 881164
650 4 $a Statistics. $3 517247
690 $a 0399
690 $a 0463
690 $a 0723
710 2 0 $a The University of North Carolina at Chapel Hill. $3 1017449
773 0 $t Dissertation Abstracts International $g 63-11A.
790 $a 0153
790 1 0 $a Losee, Robert M., $e advisor
791 $a Ph.D.
792 $a 2002
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3070861