Abstract:
Buidling off a foundation of knowledge from studies into modelling wind speed, models are
fitted to multimodal datasets of geological nature. Mixtures of distributions are derived with parameter
updates done via implementation of the EM algorithm. Among these mixtures is the
Birnbaum-Saunders which is used as a component of hierarchical mixture of multiple distributions
for the first time. The derivations of parameter updates in the EM algorithm setting is done
and application to five real world datasets, one of which is large, implemented whilst keeping
computation in mind. Additionally a simulation study is done for the mixtures of distributions
with results indicating larger samples result in better fit whilst not compromising runtimes. Simulation
studies for hierarcical mixtures to be considered in future work as obtaining convergence is
challenging.
