Tid: 1 mars 1999 kl 1515-1700
Plats : Seminarierummet 3733, Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Martin Jacobsen, Teoretisk statistik, Köpenhamns Universitet.
Titel: Discretely observed diffusions: classes of estimating functions and small delta optimality.
Sammanfattning:
Consider an ergodic diffusion process, depending on some unknown parameters, observed discretely at n equidistant point in time. It is usually very difficult - if not downright impossible - to perform exact maximum-likelihood estimation, so instead one resorts to the use of unbiased estimating functions that at least lead to estimators that are consistent and asymptotically Gaussian as n tends to infinity.
In the talk the different types of estimating functions suggested in the literature are discussed together with their asymptotic properties. The problem of finding estimating functions that are in some sense optimal will receive particular attention - e.g. through the new concept of small delta optimality.