Bean, NigelLewis, Angus2019-08-062019-08-062018http://hdl.handle.net/2440/120418Markovian-Regime-Switching (MRS) models are commonly used for modelling economic time series, including electricity prices. In this application it is common to include inde- pendent regimes as these can more accurately capture the dynamics of electricity prices compared to traditional MRS models. The advantage of independent regime MRS specifications is that they allow us to seperate dynamics between regimes. Despite their popularity, parameter inference for MRS models with independent regimes is underdeveloped. Until this thesis, there was no computationally feasible method to evaluate the likelihood of, or find maximum likelihood estimate for, MRS models with independent regimes. Moreover, there are no good discussions of Bayesian methods for such models applied to electricity prices. In this thesis we develop both maximum likelihood and Bayesian inference methodologies for MRS models with independent regimes, and use simulations to investigate their behaviours. We use our methods to investigate the South Australian wholesale electricity market, and find evidence of a significant jump in price volatility which coincides with the closure of South Australia's only coal generation facility, and therefore a significant change in market structure. Our work also suggests that Bayesian methods can be advantageous compared to maximum likelihood, since Bayesian methods can avoid issues with inferring parameters of shifted distributions, which are commonly used in this context.enRegime-Switching Time-seriesindependent regimeforward-backward algorithmexpectation maximisationdata-augmented MCMCThesis