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

Iterative logistic ridge regression.

Nichols, Sara Lyn Baty.

FindBook

Google Book

Amazon

博客來

Iterative logistic ridge regression.

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Iterative logistic ridge regression./
作者:	Nichols, Sara Lyn Baty.
面頁冊數:	78 p.
附註:	Source: Dissertation Abstracts International, Volume: 61-12, Section: B, page: 6226.
Contained By:	Dissertation Abstracts International61-12B.
標題:	Biology, Biostatistics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=9996146
ISBN:	049304860X

Iterative logistic ridge regression.
Nichols, Sara Lyn Baty.

Iterative logistic ridge regression. - 78 p.

Source: Dissertation Abstracts International, Volume: 61-12, Section: B, page: 6226.

Thesis (Ph.D.)--The University of Iowa, 2000.

Logistic regression is a well-known method for binary response modeling. The coefficient vector β is estimated using iteratively reweighted least squares (IRLS). Generally, but not always, the process converges. When the IRLS algorithm fails to converge, the maximum likelihood (ML) estimate <math> <f> <a><ac><g>b</g></ac><ac>&d4;</ac></a></f> </math> cannot be computed. Stratification and conditional maximum likelihood as implemented by the software LogXact are two common alternatives. I show, through five Monte Carlo studies, that iterative logistic ridge regression (ILRR) should be considered a reasonable alternative. ILRR applies a ridge at each iteration of the IRLS algorithm. Two adaptive ridge constants from the linear and logistic ridge literature are computed in two ways, utilizing information from the preceeding iteration or information from the current iteration. Two variance estimates, a sandwich estimate and an ad hoc estimate having the form of the ordinary ML variance estimate are explored.

ISBN: 049304860XSubjects--Topical Terms:

1018416
Biology, Biostatistics.

Iterative logistic ridge regression.
LDR:02771nam 2200301 a 45 001 929258
005 20110427
008 110427s2000 eng d
020 $a 049304860X
035 $a (UnM)AAI9996146
035 $a AAI9996146
040 $a UnM $c UnM
100 1 $a Nichols, Sara Lyn Baty. $3 1252743
245 1 0 $a Iterative logistic ridge regression.
300 $a 78 p.
500 $a Source: Dissertation Abstracts International, Volume: 61-12, Section: B, page: 6226.
500 $a Supervisor: George G. Woodworth.
502 $a Thesis (Ph.D.)--The University of Iowa, 2000.
520 $a Logistic regression is a well-known method for binary response modeling. The coefficient vector β is estimated using iteratively reweighted least squares (IRLS). Generally, but not always, the process converges. When the IRLS algorithm fails to converge, the maximum likelihood (ML) estimate <math> <f> <a><ac><g>b</g></ac><ac>&d4;</ac></a></f> </math> cannot be computed. Stratification and conditional maximum likelihood as implemented by the software LogXact are two common alternatives. I show, through five Monte Carlo studies, that iterative logistic ridge regression (ILRR) should be considered a reasonable alternative. ILRR applies a ridge at each iteration of the IRLS algorithm. Two adaptive ridge constants from the linear and logistic ridge literature are computed in two ways, utilizing information from the preceeding iteration or information from the current iteration. Two variance estimates, a sandwich estimate and an ad hoc estimate having the form of the ordinary ML variance estimate are explored.
520 $a The ridge constant <italic>k<sub>b</sub></italic> using current information and the sandwich estimate for the variance is shown, in general, to have smaller RMSE, good coverage and smaller bias relative to either the ordinary ML estimate, or to a hybrid of ordinary ML when convergent and ILRR when not. The iterative ridge estimate also compared favorably with the LogXact estimate. After adjusting for the difference in coverage the iterative ridge estimate had a shorter confidence interval than the LogXact estimate. Its relative bias was similar to and in some cases smaller than the LogXact estimate. Thus, the iterative ridge estimate performed well relative to ordinary ML estimate when the IRLS algorithm converged and it performed well relative to a hybrid estimate and to LogXact when the IRLS algorithm failed to converge for ordinary ML estimation.
590 $a School code: 0096.
650 4 $a Biology, Biostatistics. $3 1018416
650 4 $a Health Sciences, Public Health. $3 1017659
650 4 $a Statistics. $3 517247
690 $a 0308
690 $a 0463
690 $a 0573
710 2 0 $a The University of Iowa. $3 1017439
773 0 $t Dissertation Abstracts International $g 61-12B.
790 $a 0096
790 1 0 $a Woodworth, George G., $e advisor
791 $a Ph.D.
792 $a 2000
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=9996146