Executive Summary : | The author's M.Phil. dissertation explores zero-inflated integer autoregressive INAR models, including Poisson, geometric, Poisson-Lindley, and zero-one-inflated models. The dissertation focuses on statistical investigations, parameter estimation, consistency, asymptotic normality, and forecasting procedures. The proposal aims to identify research gaps in zero-inflated INAR models, including serial dependence tests, non-parametric test procedures for stationarity, randomness tests, and INARMA-type models for non-stationary data sets with zero inflation. The proposal also plans to develop coherent and Bayesian forecasts, a new class of INAR models using hyper-Poisson or alternative hyper-Poisson distributions, and machine learning algorithms for forecasting INAR models. |