Dr Martin Janzura
Institute of Information Theory and Automation, Academy of Sciences of the Czech Republic, Prague

The pseudo-likelihood principle in statistical inference

Abstract: For large stochastic systems the standard methods of statistical inference based on the maximum likelihood approach meet problems that make their implementation hardly possible. The pseudo-likelihood approach, dealing with small-dimensional conditional marginals, can avoid these fundamental difficulties. The derived methods are not strictly optimal (there is a decrease of efficiency in parameter estimation, and a slower rate of convergence in hypothesis testing), but they are good enough and the numerical feasibility is crucial.

Tid och plats måndagen den 25 mars kl 15.15 - 17.00 i seminarierum 3733, Institutionen för matematik, KTH, Lindstedtsvägen 25, plan 7.