Matstat seminar fredagen 31 augusti 13.15 i MH:227

Mogens Bladt
Mexico C and Copenhagen
    
Inference for functionals which depends on phase-type distributions.

In applied probability many functionals can be calculated explicitly or
exact numerically when the underlying distributions of the model is of
phase-type. This is for example the case for waiting time distributions 
in queues or ruin probabilities in risk theory. In many applications one
would be more interested in the statistical behaviour of the functional
itself rather than the underlying distribution(s). We establish a method
for obtaining distributions of the functionals based on data and the
phase-type assumption using Markov chain Monte Carlo. Also we compare our
method to maximum likelihood estimation of the underlying distributions.