Asymptotic Statistics and Related Topics: Theories and Methodologies

[Odalric-Ambrym Maillard]

Asymptotic Statistics and Related Topics:
Theories and Methodologies

2-4 September 2013,
Sanjo Conference Hall, The University of Tokyo, Tokyo, Japan

Bernoulli Society Satellite Meeting to the ISI World Statistics Congress 2013

Overview:
Recent developments in statistics require stronger tie-ups and integration of various research areas such as asymptotic decision theory, stochastic analysis, asymptotic distribution theory, computational statistics, stochastic numerical analysis, statistical software, machine learning, and applied fields. Topics in asymptotic statistics and related fields will be discussed comprehensively in this meeting.

Topics in asymptotic statistics and related fields will be discussed comprehensively in this conference:

1. Stochastic analysis and limit theorems
   Advanced knowledge of stochastic analysis and limit theorem is indispensable for statistical inference for stochastic-process models, which include diffusions with jumps, Lévy processes, and fractional processes, to mention just a few. Limit theorems for covariance estimators and power variations are now hot spots in high-frequency financial data analysis. Interaction between limit theorems for stochastic processes with the Malliavin calculus gives a new perspective to asymptotic distribution theory.
   
   
2. Theory of asymptotic inference and its applications
   Statistical inference for sampled stochastic processes requires new developments in asymptotic theory. Quasi likelihood analysis serves as guiding principles. Change point problems suggest importance of a framework of the asymptotic decision theory. Nonlinear time series analysis gives many problems in construction of estimating function and computations. Statistical inference for jump processes is producing various asymptotic results that are essentially different from the classical asymptotic theory. We invite talks on computational methods such as bootstrap and MCMC. We also encourage topics in statistical modeling.
   
   
3. Stochastic numerical analysis and computational statistics
   Discretization and error estimation are of importance in applications of stochastic differential equations. These are deeply related with the sampling problem in statistics for stochastic processes. Asymptotic expansion methods and their half-analytic variations are nowadays used in real financial world. Implementation is a question we should answer. Topics in computational statistics and applications to data analysis are also welcome in the session.
   
   
4. Insurance mathematics and risk theory
   Analysis of financial and insurance risks are extremely important in recent economics. Those risks have been well studied in a long history of actuarial science. After progresses in probability and mathematical statistics, insurance risk theory is dramatically changing. We will discuss recent developments from mathematical and statistical point of view.
   
   
5. Statistical learning theory
   Machine learning is one of scientific fields to which theoretical statistics makes fundamental contributions. Some of topics, e.g., reinforcement learning, ensemble learning, random forest, support vector machine, sparse coding, will be discussed. Nonparametric and semiparametric methods play an essential role. Talks in linear/nonlinear multivariate analysis, density estimation or model selection enter any one of sessions, depending on the theme.

Organizers:

Kuriki, Satoshi (Institute of Statistical Mathematics)
Lee, Sang-Yeol (Seoul National University)
Masuda, Hiroki (Kyushu University)
Shimizu, Yasutaka (Osaka University)
Suzuki, Taiji (The University of Tokyo)
Uchida, Masayuki (Osaka University)
Yoshida, Nakahiro (The University of Tokyo)

Website: http://www.sigmath.es.osaka-u.ac.jp/~tokyo2013/