Abstract:
Female Genital Mutilation (FGM) is one of harmful traditional practice in developing countries including Ethiopia. This practice causes significant and irreversible damage to the physical, psychological and sexual health of many women and girls and is one of the most devastating human rights violations. Generalized estimating equations are an extension of GLMs to accommodate correlated data . The focus of the GEE is on estimatin g the average response over the population rather than the regression parameters that would enable prediction of the effect of changing one or more independent variables on a given individual. Multilevel analysis is a methodology for the analysis of data manifesting complex variability, with a focus on the nested source of variability. Objective: The objectives of this study are to identify factors that influence female genital mutilation, the regional variability of female genital mutilation in Ethiopia , and modelling female genital mutilation using generalized estimating equations and multilevel models . Methodology: Data of Ethiopian demographic and health surveys (EDHS) of 2016 was used in this research. It includes nationally representative of 16,583 ever married women aged 15 -49. Hot-deck multiple imputations were used to handle missing in data and improve the reliability of the inference. Generalized Estimating Equations and Multilevel Logistic Regression were carried out to analyse covariates related to FGM among women and daughters’ included in the study with statistical package R. Result and Conclusion: The results obtained from the generalized estimating equation and multilevel logistic regression showed that age, type of residence, religion and education level significantly associated with female genital mutilation among women and there is a variation of female genital mutilation across region. In addition to that age of mother and circumcision status of mother’s significantly related with female genital mutilation among daughters and there is variation of female genital mutilation across the region. We compared two model (GEE and multilevel) to identify the model well describe the association of explanatory and response variables. Using standard error corresponding parameters, the multilevel analysis is used for further discussion. It was also found that a random intercept model was the best description of the data set among multilevel models.
