Author : Boby Mukta, Farjana


Accuracy Measure of Separate and Joint Modelling for a Correlated Binary Outcome: The Case Study of Mother Education and Immunization in Bangladesh

Md Asadullah; Nahid Hosen; Mamunar Rashid; Priyanka Basu; Farjana Boby Mukta; Emon Ahmed

Canadian Journal of Medicine, 2021, Volume 3, Issue Issue 3, Pages 153-161
DOI: 10.33844/cjm.2021.60514

Joint modelling is a statistical approach that is used to analyze correlated data when two or more outcome variables are correlated. By joint modeling, we refer to the simultaneous analysis of two or more different response variables from the same individual. But in a separate model, it is unable to measure the effect of covariate simultaneously. This article focuses on separate and joint modelling for correlated discrete data, including logistic regression models for binary outcomes. Since most of the women are illiterate in Bangladesh and most of the people are living in urban areas, as a result, most of the women are not aware of immunization. But an educated mother is always aware of her child's health which is dependent on immunization. Therefore, mother education and immunization are interdependent. We jointly address the effect of maternal education and immunization. Joint modeling of these two outcomes is appropriate because mother education helps raise awareness of the child's health and immunization is the prevention of various diseases for the child's health. We also identified factors influencing maternal education and immunization among women in Bangladesh. By jointly modelling we found the correlation between maternal education and immunization and the most important contributing factor. The joint model removes a less significant impact of covariates as opposed to separate models. These findings further suggested that the simultaneous impact of correlated outcomes can be adequately addressed between different responses, which is overestimated or underestimated when examined separately.