Generalized Estimating Equation (GEE) Method Approach to Determine The Best Longitudinal Regression Model on Human Development Index Data In West Java
Hilda Nurmuslimah, Asep Solih Awalluddin, Rini Cahyandari
Islamic State University of Sunan Gunung Djati Bandung
Abstract
Longitudinal regression is a statistical method used to analyze the relationship between one or more independent variables and dependent variables on longitudinal data. Longitudinal data is the result of observation of a study in several different periods. Collecting data periodically causes the correlation in each object observation so that longitudinal data could not be analyzed with the usual regression methods. One of method used for regression analysis of correlated data is Generalized Estimating Equation (GEE). GEE method provides many types of correlation structures that are independent, exchangeable, AR(1) and unstructured. This research has intention to determine the best longitudinal regression models on study case of Human Development Index (HDI) in 2010-2014 used GEE method and four different correlation structures. The best model of it will be determined by using Quasi-likelihood Information Criterion (QIC) test. The result of analysis by seeing the smallest value of QIC is AR (1).
Keywords: Longitudinal Regression,Generalized Estimating Equation,Correlation,Human Development Index
Topic: Mathematics