自動適應的無母數濾波與估計(1/2)

We study nonparametric estimation of change-points in regression functions and
parameter functions in mulitparameter likelihood models. Change-points are commonly seen in
financial and economic studies, but there has been no satisfactory methods to estimation them.
We developed kernel estimators that have optimal convergence rates. A simulation study and an
application to exchange rate data analysis indicate that our methods are very appealing for
practices. Recently, there is increasing interest in applying multiparameter likelihood models to
analyze extreme values. We proposed methods that efficiently reduce variation in nonparametric
estimation of the parameter functions. An application to annual maximum temperatures shows
that our methods significantly reduce the variation.