Workshops

Projective power entropy and detectability for hidden structures

93
reads

Shinto Eguchi

2011-12-20
13:20:00 - 14:10:00

101 , Mathematics Research Center Building (ori. New Math. Bldg.)

We discuss parametric estimation for an exponential family with considerable model uncertainty. For example, the underlying distribution may be k normal mixture distribution even if we assume a normal family. Then the maximum likelihood estimator can suggest only the unique mean; the minimum projective power entropy estimator (MPPE) can properly suggest k means as the local minima. We explore that MPPE satisfies such detectability for hidden structures beyond the model in the light of duality between statistical model and estimation.