Multilevel Modelling of Modern Contraceptive Use among Rural and Urban Population of Ethiopia

Tilahun Ferede Asena
American Journal of Mathematics and Statistics, 2013: Volume 3, Number 1, 16


The present data set has a two-level hierarchical structure, with over 16,700 women nested within eleven geographical regions in Ethiopia. The bivariate analysis result showed that Place of residence, Working status, exposure to media messages, educational status, and women religion had shown a significant variation. Accordingly, from the multilevel logistic regression model it was found that all the three models are found to be significant indicating that there is real multilevel variation among contraceptive users in Ethiopia. The deviance-based chi-square value is significant for multilevel random intercept model implies that in comparison to the model with multilevel random intercept and fixed slope model the multilevel random intercept and random coefficients model has a better fit. This further implied that multilevel logistic regression model is best fit over the ordinary multiple logistic regression models, further from the model fit diagnostics statistics, Deviance, AIC and BIC presented on Table 2. From the random coefficient estimates for intercepts and the slopes vary significantly, which implies that there is a considerable variation in the effects of religion, place of residence and radio messages, these variables also found to differ significantly across the regions. The variance component for the variance of intercept in the random slope model is large compared to its standard error. Thus, there remains some regional-level variance unaccounted for in the model. Generally, this study revealed that socio-economic, demographic and proximate variables are important factors that affect contraceptives use in Ethiopia. In line with this regional differentials shows that women in more urbanized regions such as Addis Ababa, Dire Dawa, and Harari are more likely to use modern contraceptives than respondents in regions that are more rural. The effect of regional variations for religion, place of residence and radio messages further implies that there exist considerable deference in modern contraceptive use among regions and a model with a random coefficient or slope is more appropriate to explain the regional variation than a model with fixed coefficients or without random effects. As there is variation and differences in use of Modern contraceptives across regions in Ethiopia, it is recommended to balance the effect of those factors across rural regions in Ethiopia. Researchers should use multilevel models than traditional regression methods when their data structure is hierarchal as like in EDHS data.

Number of levels
Model data structure
Response types
Multivariate response model?
Longitudinal data?
Further model keywords
Substantive discipline
Substantive keywords
Paper submitted by
Tilahun Ferede Asena, Statistics, Arba Minch University,
Edit this page