A BAYESIAN SIMULATION APPROACH TO MODELING THE RELATIONSHIP BETWEEN NARCOTIC DRUG USE PREVALENCE AND UNEMPLOYMENT RATE USING AGGREGATE DATA

Abstract

Fitting a Binary Logistic Model relating narcotic drug use prevalence and unemployment rate can be a challenge in the face of aggregate data. This aggregation limits the use of the Classical Binary Logistic Regression Model. This limitation informed our development of a Bayesian Statistical Simulation Modeling Procedure. The modeling procedure embeds the Markov Chain Monte Carlo (MCMC) algorithm which is implemented on an Open Source Software Platform-Windows; Bayesian Inference Using the Gibbs Sampler (WINBUG). Part of the modeling activity is the mathematical analysis on the response of the success probability of narcotic drug use prevalence to changes in unemployment rate. This was done under conditions of positive and negative values of the regression parameters (constant and covariate coefficient). The extent to which unemployment rate is a risk factor of the success probability was investigated and the zones within Nigeria where unemployment rate is a risk factor of narcotic drug use prevalence where also identified. Results revealed that unemployment rate is a risk factor to narcotic drug use prevalence in four (4) zones (North-Central, South-East, South-West and North-East) with risk levels of 53.65%, 51.59%, 49.42% and 46.02% respectively. While, factors latent to study impact negatively on five (5) zones (North-East, North-West, North-Central, South-East and South-West). It is recommended that attention should be drawn to the South-South zone where unemployment rate is not a risk factor to narcotic drug use prevalence, but other factors latent to the study are impacting positively on it.