Multilevel Logistic Modeling of Adult Mortality Among Regional States in Ethiopia

Abstract

The purpose of this study is to investigate adult mortality variations among regional states in Ethiopia. Data from Ethiopian Demographic and Health Survey (EDHS 2011) are used. Multilevel logistic regression model is fitted to the data. The findings show that the correlates of adult mortality are adult's age, age of household head, sex of household head, wealth index, type of toilet facility, and type of cooking fuels. The between-region variance is estimated to be 0.1727 which is significant at 5% level of significance, indicating that variation of adult mortality among regional states is non-zero. The intra-correlation is estimated to be 0.0099, suggesting that about 1% of the variation in adult mortality could be attributed to differences across regional states. The variance of the random coefficient model is statistically significant, implying the presence of adult mortality variation among regional states of the country. The deviance statistic is about 4113.945 and it is compared with Chi-square at difference of full and empty model with degree of freedom 33.924 at df 22 and p-value < 0.000. This reveals that there is enough evidence against null hypothesis. Therefore, the multilevel model consisting the explanatory variables is considered the final model. It revealed that the random coefficient multilevel logistic regression analysis suggests that there exist considerable differences in adult mortality across regions in the country.