Seemingly Unrelated Exponetiated Exponential Geometric Regression Model

In the class of seemingly unrelated regression models, the dispersion nature of the dependent variable can greatly impact the efficiency and reliability of the parameter estimates for the model. Despite this, the seemingly unrelated Poisson regression model and seemingly unrelated negative binomial model are two most commonly used count data models for these class of regression models. This study introduces the seemingly unrelated exponentiated exponential geometric regression (SUEEGR) for modelling count data which might be equi-, under, or over-dispersed. Parameters estimation for the model was carried out using the method of maximum likelihood. A simulation study was carried out to assess the performance of the model under various conditions using certain evaluation criteria for the under-dispersed, over dispersed and when SUEEGR reduces to SU-geometric regression. The model was applied to datasets from Demography and Health Survey (DHS), the descriptive with scattered plot of the dataset were obtained and the results was compared with those obtained from seemingly unrelated Poisson regression. From the findings, it was observed that there is a significant correlation (0.1594) between the response variables and SUEEGR model appears to provide a better overall fit of the data than SUGPR because of the higher Log-likelihood, lower AIC and BIC of the model.