Journal of Big Data (Apr 2021)
Bayesian negative binomial logit hurdle and zero-inflated model for characterizing smoking intensity
Abstract
Abstract Smoking invariably has environmental, social, economic and health consequences in Ethiopia. Reducing and quitting cigarette smoking improves individual health and increases available household funds for education, food and better economic productivity. Therefore, this study aimed to apply the Bayesian negative binomial logit hurdle and zero-inflated model to determine associated factors of the number of cigarette smokers per day using the smoking intensity data of 2016 Ethiopia Demographic and Health Survey. The survey was a community-based cross-sectional study conducted from January 18 to June 27, 2016. The survey used two stage stratified sampling design. Bayesian analysis of Negative Binomial Logit Hurdle and Zero-inflated models which incorporate both overdispersion and excess zeros and carry out estimation using Markov Chain Monte Carlo techniques. About 94.2% of them never cigarettes smoked per day and the data were found to have excess zeros and overdispersion. Therefore, after considering both the zero counts and the enduring overdispersion, according to the AIC and Vuong tests, the Zero-inflated Negative Binomial and Negative Binomial Logit Hurdle model best fit to the data. The finding Bayesian estimation technique is more robust and precisely due to that it is more popular data analysis method. Furthermore; using Bayesian Zero-inflation and Zero hurdle model the variable: age, residence, education level, internet use, wealth index, marital status, chewed chat, occupation, the media were the most statistically significant determinate factors on the smoking intensity.
Keywords