Generalized estimating equation (GEE) on binary longitudinal data

Abstract. Binary logistic regression is a regression in which the response variable is binary with one or more predictor variables being both categorical and continuous. It can be applied to longitudinal data, where the observations are cross-sectional units over several periods of time. These repeated observations cause autocorrelation and needs to be addressed by implementing the Generalized Estimating Equation (GEE) method. The study aims to apply GEE and choose the best correlation structure based on the QIC value, where the data is the sputum status of pulmonary tuberculosis patients at PKU Muhammadiyah Hospital at Bantul, Yogyakarta.  Based on the analysis then the model of sputum status is , with the correlation structure is autoregressive (1).