Bivariate ordered probit modelling of motorcycle riders and pillion passengers' injury severities relationship and associated risk factors.

This study simultaneously modelled the injury severity of motorcycle riders and their pillion passengers and determine the associated risk factors. The analysis is based on motorcycle crashes data in Ashanti region of Ghana spanning from 2017 to 2019. The study implemented bivariate ordered probit model to identify the possible risk factors under the premise that the injury severity of pillion passenger is endogenously related to that of the rider in the event of crash. The model provides more efficient estimates by considered the common unobserved factors shared between rider and pillion passenger. The result shows a significant positive relationship between the two injury severities with a correlation coefficient of 0.63. Thus, the unobservable factors that increase the probability of the rider to sustain more severe injury in the event of crash also increase that of their corresponding pillion passenger. The rider and their pillion passenger injury severities have different propensity to some of the risk factors including passengers' gender, day of week, road width and light condition. In addition, the study found that time of day, weather condition, collision type, and number of vehicles involved in the crash jointly influence the injury severity of both rider and pillion passenger significantly.

Modelling COVID-19 incidence in the African sub-region using smooth transition autoregressive model.

Prediction of COVID-19 incidence and transmissibility rates are essential to inform disease control policy and allocation of limited resources (especially to hotspots), and also to prepare towards healthcare facilities demand. This study demonstrates the capabilities of nonlinear smooth transition autoregressive (STAR) model for improved forecasting of COVID-19 incidence in the Africa sub-region were investigated. Data used in the study were daily confirmed new cases of COVID-19 from February 25 to August 31, 2020. The results from the study showed the nonlinear STAR-type model with logistic transition function aptly captured the nonlinear dynamics in the data and provided a better fit for the data than the linear model. The nonlinear STAR-type model further outperformed the linear autoregressive model for predicting both in-sample and out-of-sample incidence.