Parametric models for random polyhedral grains and statistics in case of incomplete data

Felix Balani (Univ. Freiberg)


Abstract:

For a parametric model for random polyhedral grains which is given as an exponential family we discuss how parameters 
may be estimated in case not all minimal sufficient statistics are available. Due to the nature of the model any kind 
of parameter estimation needs to be based on Monte Carlo which makes even the usually particularly suitable 
expectation-maximization algorithm practically not feasible. It turns out that instead an approach based on the 
quasi-likelihood theory can be successfully applied.