The assumptions used for modelling the demand for blood vary widely in the literature. In this talk, a case is presented that blood transfusions are best modelled using an acyclic discrete phase type distribution (ADPH). However, in the general case the ADPH is over specified with complexity O(n2) . It is shown how a general ADPH can be transformed into any one of three canonical forms each of which is a unique minimum representation with complexity O(n). Using one of these forms a Gibbs sampler used to generate credible parameter estimates of the ADPH. This in turn requires time reversal on a non-stationary Markov chain in order to generate credible sample paths on the ADPH. The whole process is demonstrated with an application to real blood transfusion data.
About the speaker
Nigel Clay is a third year IDTC PhD candidate in Mathematical Science at RMIT University. His industry partner is the Australian Red Cross Blood Service and consequently his research considers inventory issues in the blood supply chain. This covers system dynamics, modelling and estimation of discrete phase-type distributions and discrete time stochastic control.
How to participate in this seminar
(No access to an ACE facility? Contact Maaike Wienk for a Visimeet guest licence – limited licences available – first come first serve)