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

Analysis of stratified longitudinal ...

Swann, Rhonda Suzanne.

FindBook

Google Book

Amazon

博客來

Analysis of stratified longitudinal studies using generalized estimating equations with data missing at random.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Analysis of stratified longitudinal studies using generalized estimating equations with data missing at random./
作者:	Swann, Rhonda Suzanne.
面頁冊數:	314 p.
附註:	Source: Dissertation Abstracts International, Volume: 63-05, Section: B, page: 2147.
Contained By:	Dissertation Abstracts International63-05B.
標題:	Biology, Biostatistics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3052078
ISBN:	0493664270

Analysis of stratified longitudinal studies using generalized estimating equations with data missing at random.
Swann, Rhonda Suzanne.

Analysis of stratified longitudinal studies using generalized estimating equations with data missing at random. - 314 p.

Source: Dissertation Abstracts International, Volume: 63-05, Section: B, page: 2147.

Thesis (Ph.D.)--University of South Carolina, 2002.

This work investigated methods to handle missing covariate and response data in stratified, longitudinal studies analyzed using generalized estimating equations (GEE). Public health studies often use complex sampling to maximize information and minimize costs when surveying human populations. Analysis of data from complex samples most correctly uses design factors such as the size of the finite population, sampling probabilities, and sample size.

ISBN: 0493664270Subjects--Topical Terms:

1018416
Biology, Biostatistics.

Analysis of stratified longitudinal studies using generalized estimating equations with data missing at random.
LDR:03175nmm 2200313 4500 001 1836733
005 20050315121040.5
008 130614s2002 eng d
020 $a 0493664270
035 $a (UnM)AAI3052078
035 $a AAI3052078
040 $a UnM $c UnM
100 1 $a Swann, Rhonda Suzanne. $3 1925201
245 1 0 $a Analysis of stratified longitudinal studies using generalized estimating equations with data missing at random.
300 $a 314 p.
500 $a Source: Dissertation Abstracts International, Volume: 63-05, Section: B, page: 2147.
500 $a Major Professor: Cheryl L. Addy.
502 $a Thesis (Ph.D.)--University of South Carolina, 2002.
520 $a This work investigated methods to handle missing covariate and response data in stratified, longitudinal studies analyzed using generalized estimating equations (GEE). Public health studies often use complex sampling to maximize information and minimize costs when surveying human populations. Analysis of data from complex samples most correctly uses design factors such as the size of the finite population, sampling probabilities, and sample size.
520 $a Regardless of the sampling design, missing data are common. For data missing at random (MAR), the probability of an observation being missing is conditionally independent of the unobserved data given the observed data. This study investigates MAR covariates and responses when the data are a binary longitudinal outcome and two baseline binary covariates using samples from simulated stratified finite populations.
520 $a For an MAR covariate, weighted generalized estimating equations (WGEE) were compared to complete case analysis when using GEE incorporating the sampling probabilities. The longitudinal response and the other covariate were fully observed. For MAR longitudinal responses with both covariates fully observed, modified generalized estimating equations (MGEE) and model-based multiple imputation assuming a multivariate normal distribution of the missing responses were compared to complete case analysis. Results were assessed using bias, coverage probabilities, and calculated and empirical variance estimates.
520 $a WGEE with a correct or over-specified model predicting missingness of the covariate performed better than either complete case analysis or WGEE using an under-specified model. Parameter estimates from MGEE with intermittently missing responses are best when the assumed correlation structure of the analysis is correct, and MGEE consistently outperformed complete case analysis. The MI method produced severely biased estimates compared to complete case analysis. It is more sensitive to the imputation model than to the assumed correlation. The coefficient point estimates had less variability than indicated by calculated variance in all three methods, which could lead to a loss of power to detect weak associations.
590 $a School code: 0202.
650 4 $a Biology, Biostatistics. $3 1018416
650 4 $a Health Sciences, Public Health. $3 1017659
690 $a 0308
690 $a 0573
710 2 0 $a University of South Carolina. $3 1017477
773 0 $t Dissertation Abstracts International $g 63-05B.
790 1 0 $a Addy, Cheryl L., $e advisor
790 $a 0202
791 $a Ph.D.
792 $a 2002
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3052078