2018 Fiscal Year Final Research Report
Efficient Use of a Weakly Informative Prior: Bayesian Likelihood and the Unification of Multiple Sources of Information
| Project/Area Number |
15K00061
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Statistical science
|
|---|
| Research Institution | The Institute of Statistical Mathematics |
|---|
Principal Investigator |
|
|---|
| Research Collaborator |
Ogura Toru
Tahata Kouji
Ohkusa Kosuke
|
|---|
| Project Period (FY) |
2015-04-01 – 2019-03-31
|
| Keywords | Bayesian theory / Estimation / e-divergence / Informative prior |
|---|
| Outline of Final Research Achievements |
The proposed research subject was to obtain properties of the e-mixture Bayesian likelihood and its implications to the unification of multiple sources of information. This task was performed in a satisfactory way. This success is fortunately supported by the fact that Jeffreys' or the reference prior implies favorable estimators. In other words, the maximum likelihood estimator behaves poorly. In the last scheduled yea necessary risk comparison studies were conducted.
In this process we obtain a new idea of our further studies. The Laplace distribution has been attracted reseachers' attentions, because of its favorable relationship between the median and the maximum likelihood estimator. It becomes clear that our approach can be extended to covering this distribution, which allows us the efficient use of this promising distribution.
|
| Free Research Field |
Theoretical Statstics
|
| Academic Significance and Societal Importance of the Research Achievements |
本研究は、今日に必要とされる証拠に基づいた医療・行政などを支える基礎技術の発展を支える。別の面から見るとビッグデータの収集方法についての議論についても精密なデータが筆であるとの示唆を与える。但し、現在の成果が直接に役立つレベルに達していない。
|
