• ARAGAW ESHETIE AGUADE
• Prof. B.Muniswamy

Overdispersion is familiar in count data models particularly in the area of ecological and biological science because of non- independent, aggregations of data and an excess frequency of zeros. Every cluster levels received a singular level of a random effect that models the additional Poisson variation given within the data, are usually utilized to address overdispersion in count data. However, studies investigating that the power of cluster level random effects as a way to data with overdispersion is scarce. A situation where the variance of the response variable exceeds the mean, and hence, both overdispersion and heterogeneity between groups parameters occur. In the appropriate imposition of the multilevel Poisson model may underestimate the standard error and overestimate the significance of the regression parameters, and consequently, giving misleading inference about the regression parameters. This paper suggested that the multilevel negative binomial models as alternatives for handling overdispersion; an algorithm is developed for the residual maximum likelihood estimate (REML) of the regression coefficients and variance component parameters. In addition, the predicted random effects can provide information on the interregional variation after adjustment for children characteristic and death features. In this paper we used an application and simulation study, a simulation study showed that the estimators obtained from multilevel negative binomial model perform well in all the setting considered. The simulations reveal that failing to account for overdispersion in mixed models can erroneously inflate measures of explained variance ( ), which may lead to researchers overestimating the predictive power of variables of interest. Application to a set of deaths of children under 18 year’s data is illustrated. The result revealed that the proposed model is better than the multilevel Poisson regression model and hence, there is a variation among regions in the deaths of children less than 18 years. Both the predicted probability and the information criteria indicated that the multilevel negative binomial model is better than the multilevel Poisson regression model. This work suggests the use of group-level random effects provides a simple and robust means to a count data, but also that this ability to minimize bias is not uniform across all types of overdispersion and must be applied thoughtfully.

Clustered count data, Over-dispersion, MZIP, MZINB, Homogeneity.

"Proposed Score Test for Over-dispersion Parameter in the Multilevel Negative Binomial Regression Model", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.5, Issue 12, page no.709-720, December-2018, Available :http://www.jetir.org/papers/JETIR1812699.pdf

"Proposed Score Test for Over-dispersion Parameter in the Multilevel Negative Binomial Regression Model", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.5, Issue 12, page no. pp709-720, December-2018, Available at : http://www.jetir.org/papers/JETIR1812699.pdf