Honours Project Gets Prize from Actuarial Society
Trevor Brinkman, a UCT Actuarial Science graduate, received the runner up prize in the Actuarial Society Project Competition for his honours project conducted under the supervision of Prof Linda Haines (Statistical Sciences). His project was entitled "The Autoregressive Conditional Poisson (ACP) and Double Autoregressive Conditional Poisson (DACP) models".
A study of the Autoregressive Conditional Poisson (ACP) and Double Autoregressive Conditional Poisson (DACP) models is undertaken in this paper. The author found that these models can be used to address the problem of overdispersion often found in integer-valued time series and also allow estimation to be done by conditional maximum likelihood. In the study, simulations are carried out in order to: (i) check theoretical expressions derived by a number of authors for the mean and variance, (ii) investigate the accuracy and precision of the conditional Maximum Likelihood Estimates and (iii) check theoretical results derived for the forecasts of these models along with the accuracy of forecast confidence intervals. These models are then applied to two real data sets, from which the various models are compared with regard to their forecasting ability as based on a number of factors. When comparing the ACP and DACP model frameworks, the author found that the ACP model performs better at making point forecasts while the DACP model produced better forecast confidence intervals.