Abstract
Adolescence behavior is a very crucial factor that influences the present and future health and development. The aim of this study was to select the appropriate model to analyze the adolescent behavior in Bangladesh. Follow-up data which consisting of five consecutive rounds during the 30 months period from January 2008 to June 2010 has been used for this study. Generalized estimating equations (GEE) models with independence, exchangeable, and unstructured correlation structures was performed to analyze the data. Based on the modified version of Akaike's Information Criterion (mAIC), independence correlation structure exhibits better results to analyzing the adolescent behavior with covariates education of adolescent, sex, marital status, occupation and number of session attended.