Yarı En Küçük Kareler Regresyonu Yönteminin Farklı Çalışan Korelasyon Yapılarına İlişkin Sonuçların Karşılaştırılması ve Sağlık Alanında Uygulaması
![](/xmlui/themes/Mirage2//images/mime2.png)
View/ Open
Date
2019Author
ASAR, Erdoğan
xmlui.dri2xhtml.METS-1.0.item-emb
Acik erisimxmlui.mirage2.itemSummaryView.MetaData
Show full item recordAbstract
In the present study, introducing of the Quasi-Least Squares Regression (QLS) which is obtained of the extension of the Generalized Estimating Equations (GEE) and comparison of the results got from the applications of QLS and GEE are aimed. QLS is not widely used in our Country. It is a method for associated data analysis based on a two-stage computational approach. In the context of these objectives, the real data related health area and the data obtained from a simulation study are analyzed, and the results of these studies are evaluated. In the study of the real data; the impact of the time and group variables on four dependent variables regarding 114 observations got 38 patients received orthodontics treatment is investigated. In the simulation study; 9 data set which are repeated 1000 times for three current correlation structures and three different values are generated. The results of 36 cases obtained from applying four working correlation structures on per data set are evaluated. In accordance with the simulation work: Generally, in terms of the efficiency of the estimations, QLS is superior to GEE; to compare better the results of QLS and GEE, the Markov correlation structure for GEE should be applicable as QLS. According to the real data study: The highest correlation value is obtained from the Markov working correlation structure among the working correlation structures used in QLS; the effect of the methods of maxillofacial surgery, as defined group variable, on all of the dependent variables are not significant. In both of the real work and simulation study: At the execution of the tri-diagonal working correlation structure, while the convergence is not achieved in GEE, it is achieved in QLS.