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

FindBook

Google Book

Amazon

博客來

Marginal Modeling and Multiple Imputation Methods for Correlated Binary Data Generated from Clinical Periodontal Examinations.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Marginal Modeling and Multiple Imputation Methods for Correlated Binary Data Generated from Clinical Periodontal Examinations./
作者:	Shing, Tracie L.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2022,
面頁冊數:	162 p.
附註:	Source: Dissertations Abstracts International, Volume: 83-11, Section: B.
Contained By:	Dissertations Abstracts International83-11B.
標題:	Biostatistics. -
ISBN:	9798438797630

Marginal Modeling and Multiple Imputation Methods for Correlated Binary Data Generated from Clinical Periodontal Examinations.
Shing, Tracie L.

Marginal Modeling and Multiple Imputation Methods for Correlated Binary Data Generated from Clinical Periodontal Examinations. - Ann Arbor : ProQuest Dissertations & Theses, 2022 - 162 p.

Source: Dissertations Abstracts International, Volume: 83-11, Section: B.

Thesis (Dr.P.H.)--The University of North Carolina at Chapel Hill, 2022.

This item must not be sold to any third party vendors.

Large epidemiological studies of periodontitis often use partial mouth oral examinations, where a fraction of teeth or tooth sites are examined for periodontal disease, in lieu of full mouth examinations to reduce physical burdens and costs. Partial mouth examinations consistently underestimate prevalence when case definitions classify individuals as having periodontitis using only information from observed sites. Meanwhile, the Preisser et al. (2017) prevalence formula based on simple site-level disease assumptions produces less biased prevalence estimates for partial mouth examinations; however, the estimator is generally biased as it does not incorporate complex intraoral disease patterns or case definitions.To better characterize intraoral patterns of disease, we employ marginal regression analyses in an epidemiological investigation of tooth site-level periodontal disease in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL). Since few methods and software are available for marginal analysis of binary outcomes when complex within-cluster correlation structures are of interest, we first develop a SAS macro, GEECORR, to implement the Prentice (1988) extension to the Liang and Zeger (1986) generalized estimating equations (GEE) method. This allows for flexible modeling of both the marginal mean and within-cluster correlation as a function of their respective covariate risk factors. Then, we use the Prentice (1988) GEE method to examine several mean and correlation models and detail both site-level probabilities and correlation patterns of periodontal disease in the HCHS/SOL population.Finally, we describe a novel methodology for multiply imputing missing correlated binary observations from incomplete clusters using the Prentice (1988) GEE approach to estimate an imputation model and the conditional linear family of statistical distributions Qaqish (2003) to generate "complete" clusters. We apply our imputation-based method to partial mouth examination protocols generated with simple random samples of tooth sites from the 2013-2014 National Health and Nutrition Examination Survey full mouth periodontal examination. Prevalence estimates from multiply imputed clusters are compared to gold standard full mouth estimates. We demonstrate that periodontitis prevalence estimands based on complex disease definitions for the full mouth can be validly estimated from partial mouth samples using multiple imputation to generate complete clusters.

ISBN: 9798438797630Subjects--Topical Terms:

1002712
Biostatistics.
Subjects--Index Terms:

Correlated binary data

Marginal Modeling and Multiple Imputation Methods for Correlated Binary Data Generated from Clinical Periodontal Examinations.
LDR:03658nmm a2200361 4500 001 2349678
005 20221003074928.5
008 241004s2022 eng d
020 $a 9798438797630
035 $a (MiAaPQ)AAI28869260
035 $a AAI28869260
040 $a MiAaPQ $c MiAaPQ
100 1 $a Shing, Tracie L. $3 3689096
245 1 0 $a Marginal Modeling and Multiple Imputation Methods for Correlated Binary Data Generated from Clinical Periodontal Examinations.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2022
300 $a 162 p.
500 $a Source: Dissertations Abstracts International, Volume: 83-11, Section: B.
500 $a Advisor: Preisser, John S.
502 $a Thesis (Dr.P.H.)--The University of North Carolina at Chapel Hill, 2022.
506 $a This item must not be sold to any third party vendors.
520 $a Large epidemiological studies of periodontitis often use partial mouth oral examinations, where a fraction of teeth or tooth sites are examined for periodontal disease, in lieu of full mouth examinations to reduce physical burdens and costs. Partial mouth examinations consistently underestimate prevalence when case definitions classify individuals as having periodontitis using only information from observed sites. Meanwhile, the Preisser et al. (2017) prevalence formula based on simple site-level disease assumptions produces less biased prevalence estimates for partial mouth examinations; however, the estimator is generally biased as it does not incorporate complex intraoral disease patterns or case definitions.To better characterize intraoral patterns of disease, we employ marginal regression analyses in an epidemiological investigation of tooth site-level periodontal disease in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL). Since few methods and software are available for marginal analysis of binary outcomes when complex within-cluster correlation structures are of interest, we first develop a SAS macro, GEECORR, to implement the Prentice (1988) extension to the Liang and Zeger (1986) generalized estimating equations (GEE) method. This allows for flexible modeling of both the marginal mean and within-cluster correlation as a function of their respective covariate risk factors. Then, we use the Prentice (1988) GEE method to examine several mean and correlation models and detail both site-level probabilities and correlation patterns of periodontal disease in the HCHS/SOL population.Finally, we describe a novel methodology for multiply imputing missing correlated binary observations from incomplete clusters using the Prentice (1988) GEE approach to estimate an imputation model and the conditional linear family of statistical distributions Qaqish (2003) to generate "complete" clusters. We apply our imputation-based method to partial mouth examination protocols generated with simple random samples of tooth sites from the 2013-2014 National Health and Nutrition Examination Survey full mouth periodontal examination. Prevalence estimates from multiply imputed clusters are compared to gold standard full mouth estimates. We demonstrate that periodontitis prevalence estimands based on complex disease definitions for the full mouth can be validly estimated from partial mouth samples using multiple imputation to generate complete clusters.
590 $a School code: 0153.
650 4 $a Biostatistics. $3 1002712
650 4 $a Dentistry. $3 828971
650 4 $a Epidemiology. $3 568544
653 $a Correlated binary data
653 $a Marginal models
653 $a Multiple imputation
653 $a Periodontal disease
653 $a Periodontitis
690 $a 0308
690 $a 0567
690 $a 0766
710 2 0 $a The University of North Carolina at Chapel Hill. $b Biostatistics. $3 1023527
773 0 $t Dissertations Abstracts International $g 83-11B.
790 $a 0153
791 $a Dr.P.H.
792 $a 2022
793 $a English