Faculty Profiles Main Page Publications Japanese
Professor Website
KUBOKAWA, Tatsuya Personal Website
Office E-mail
Faculty of Economics,
The University of Tokyo,
7-3-1 Hongo, Bunkyo-ku, Tokyo, 113-0033
Research Field
Mathematical Statistics
Research Theme
In the field of statistical estimation, I have developed new theories to construct useful procedures which are more efficient than usual procedures such as unbiased or maximum likelihood estimators. For small sample, these usual procedures sometimes have drawbacks of instability or inefficiency. Also they sometimes take undesirable values beyond the parameter spaces. In the estimation problems with such undesirable properties, I have constructed classes of estimators which have uniformly smaller risks than usual estimators and eliminate their shortcoming. Out of the classes, I have singled out empirical Bayes and generalized Bayes estimators and other new and useful estimators. Especially I succeeded not only in establishing a new unified theory in the classical problem of estimating variance components in mixed linear models, but also in extending the theory to multivariate variance components models. Related to this multivariate extension, an innovative method for constructing minimax estimators of a covariance matrix was obtained in multivariate linear regression models