The global impact of COVID-19 incidence and mortality propelled world leaders under the guid- ance of the World Health Organization to agree on strategies and goals to reduce infections and mortality. The Southern African region was more vulnerable to the disease due to the high preva- lence of comorbidities. Hence, it is crucial to estimate incidence, mortality, and administered vaccinations to continuously monitor this disease. This estimation generally requires a suitable distribution. Failure to find a suitable distribution can result in wrong inferences. Thus, this research aimed to identify generalized distributions that can best fit and model the COVID-19 data in the Southern African region. Graphical and numerical summaries were used to determine the most suitable probability distributions for modelling COVID-19 data in Southern Africa. Two generalized distributions and their sub-models, as well as the two classical distributions, were evaluated to estimate parameters and identify the most suitable distribution for describ- ing the characteristics of COVID-19 data in the region. The maximum likelihood estimation method was used to estimate the unknown parameters of the models. The results indicated that the Type I Heavy-Tailed Log-Logistic distribution provided a better fit for COVID-19 weekly incidence, mortality, and administered vaccinations in Southern Africa and at the seasonal level compared to its competitors. However, the Topp-Leone Half Logistic-Odd Burr X-Log-Logistic distribution o↵ered a better fit for incidence in the spring period. This study provides valuable insights in modelling heavy-tailed/non-monotonic distributions such as the COVID-19 datasets used in this research.

Thesis (MSc in Statistics)---Botswana International University of Science and Technology, 2024