[01] Ramesha Jayasinghe, Department of Statistics, University of Colombo, Colombo, Sri Lanka. [02] Roshini Sooriyarachchi, Department of Statistics, University of Colombo, Colombo, Sri Lanka.

The field of modelling, multilevel data is a new approach. This research study examines the emerging role of modelling multilevel data in the context of analysing the factors associated with number of deaths due to road traffic accidents and type of road user which has the highest death rate. One of the objectives of this project is to perform a missing value imputation in the context of multilevel data. It was successfully obtained by performing multiple imputation using ‘jomo’ package in R statistical software. Generalized linear mixed models (GLMM) within the ‘Glimmix’ procedure of ‘SAS’ software was used to model the number of road deaths response and type of road user which has the highest death rate response. The study was based on data which were retrieved from the “GLOBAL STATUS REPORT ON ROAD SAFETY 2015” which was published by World Health Organization. It consists of worldwide data related to socioeconomic, health and law variables in 180 United Nations countries in six regions. This study showed that the modelling of the number of road deaths and type of road user which has the highest death rate could be adequately done using a GLMM with a Negative Binomial model and Multinomial model respectively. A cluster effect was assumed within regions. The internal and external validation showed that the model predicts well.

Generalized Linear Mixed Model, Negative Binomial Distribution, Multinomial Distribution, SAS

[01] WHO, “GLOBAL STATUS REPORT ON ROAD SAFETY,” World Health Organization, 2015. [02] S. Gopalakrishnan, “A Public Health Perspective of Road Traffic Accidents,” Journal of family Medicine and Primary Care, vol. 1, no. 20, pp. 144-150, 2012. [03] A. T. Becerra, X. L. Bravo and I. F. Parra, “National and Regional Analysis of Road Accidents in Spain,” Traffic Injury Prevention, vol. 14, no. 5, pp. 486-495, 2013. [04] Eurostat, “Road accidents fatalities-statistics by type of vehicle,” 2015. [05] D. Meyer, A. Zeileis and K. Hornik, “Residual-based Shading for Visualizing (Conditional) Independence,” Journal of Computational and Graphical Statistics, vol. 16, no. 3, pp. 507-525, 2007. [06] D. B. De Silva and M. R. Sooriyarachchi, “Generalized Cochran Mantel Haenszel test for multilevel correlated categorical data: an algorithm and R function,” Journal of the National Science Foundation of Sri Lanka, vol. 40, no. 2, pp. 137-148, 2012. [07] J. R. Carpenter and M. G. Kenward, Multiple Imputation and its Application, A John Wiley & Sons. [08] J. Zhang and D. D. Boos, “Generalized Cochran-Mantel-Haenszel Test Statistics for correlated categorical data,” Communications in Statistics - Theory and Methods, pp. 1813-1837, 1997. [09] D. Collett, Modelling Binary Data, London: Chapman & Hall, 1991. [10] Institute for Digital Research and Education, “Introduction to Generalized Linear Mixed Models,” 10 January 2018. [Online]. Available: https://stats.idre.ucla.edu/other/mult-pkg/introduction-to-generalized-linear-mixed-models/. [11] S. M. Fernando and M. R. Sooriyarachchi, “Bivariate Negative Binomial Modelling of Epidemiological Data,” Open Science Journal of Statistics and Application, pp. 47-57, 2018. [12] D. D. B. Trinidade, R. Ospina and L. D. Amorim, “Choosing the right strategy to model longitudinal count data in Epidemiology: An application with CD4 cell counts,” Epidemiology Biostatistics and Public Health, vol. 12, no. 4, 2015. [13] A. Agresti, Categorical Data Analysis, Florida: A John Wiley & Sons, 2002. [14] SAS Inc. Institute, SAS/STAT 9. 2 User's Guide, Second Edition, Cary: SAS Pub, 2009. [15] WHO, “GLOBAL STATUS REPORT ON ROAD SAFETY,” World Health Organization, 2013. [16] H. Akaike, “A New Look at the Statistical Model Identification,” IEEE Transaction on Automatic Control, pp. AC–19, 716–723., 1974.