Generalized estimating equations approach in the case of existence a covariate in categorical data analysis and an application

ABSTRACT PhD Thesis GENERALIZED ESTIMATING EQUATION APPROACH IN THE CASE OF EXISTENCE A COVARIATE IN CATEGORICAL DATA ANALYSIS AND AN APPLICATION BERNA YAZICI Anadolu University Graduate School of Natural and Applied Sciences Statistics Program Supervisor: Prof. Dr. Embiya A?AO?LU 2001 In this thesis statistical processes which are used in the case of having a covariate in categorical data analysis are examined. First of all analysis of covariance for continiuous data sets is briefly given, then randomized model methods, weighted least squares metod, unweighted least squares method and loglinear model methods are discussed. Then information about longitudinal data sets is given and a kind of longitudinal data sets, crossover designs are mentioned. "Generalized Estimating Equations" method which is used to increase efficiency for this kind of data sets is discussed. The comparisons of Generalized Estimating Equations method with maximum likelihood method and weighted least squares method are briefly given. Finally this method is applied on a group of osteoarthritis patients data and the results are interpreted. Keywords: Covariate, Categorical Data Analysis, Generalized Estimating Equations (GEE), Analysis of Covariance