Workshops

The Applied Statistics Workshop 2005

21世紀COEプログラム「市場経済と非市場機構との連関研究拠点」と、日本経済国際共同研究センター(CIRJE)の共催ワークショップ

※ 2006年1月13日現在の予定です。

今後の予定:

日時

1月20日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

樋口知之(情報・システム研究機構、統計数理研究所)

逐次データ同化手法とその応用

要旨(Abstract)

データ同化とは,数値物理シミュレーションモデルに含まれる変数を物理モデルと観測データの両方をなるべく満足するよう修正する手法であり,近年海洋学・気象学において研究が進んでいる.物理シミュレーションモデルが含む変数の次元は,数百次元から100万次元程度の超大規模な問題であり,データ同化の研究はいろいろな意味で計算限界への挑戦である.データ同化は逐次型と非逐次型の方法があり,逐次型においてはアンサンブルカルマンフィルタなどが主に用いられているが,粒子フィルタの適用もわずかながらなされてきている.一方非逐次型においては,4 次元変分法という方法が用いられている.本発表においては,データ同化について,特に,逐次型の手法を時系列解析の枠組で紹介する.さらに実際のデータ同化の応用事例として,JST CREST研究プロジェクトで現在遂行している 1) エルニーニョ現象の解明に向けた大気海洋結合シミュレーションモデルと人工衛星による海面高度リモートセンシングデータのデータ同化,及び 2) 津波シミュレーションモデルと潮位データのデータ同化について紹介する. 最後に,データ同化手法のあらゆる分野への応用可能性と将来性を議論 したい.

 


本年度終了分:

日時

4月15日(金 Friday) 16:50-18:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Lecture Hall No.3 on the 3rd floor of the New Economics Building

報告

小林正人 Masato Kobayashi (横浜国立大学)

Tests for Jumps in Returns and Volatility of Stochastic Volatility Models

Abstract This paper proposes the Lagrange multiplier (LM) test to detect jumps in returns in the case where the null hypothesis is the simple stochastic volatility (SV) model without jumps and the alternative hypothesis is the SV model with jumps in return. It is shown that the null distributions of the test statisitc is independent of jump probability, which is unidentified under the null hypothesis. Then this test is free from the problem suggested by Davies (1977) caused by the presence of unidentified parameter under the null.
    The LM test to detect jumps in volatility is also proposed in the case where jumps in re-turns and volatility are contemporaneous and correlated under the alternative hypothesis. The test statistic is derived by regarding the degenerate density of volatility jumps as Dirac's delta function under the null hypothesis of the SV model with jumps only in returns. It is shown that the null distribution of the test statistic is independent of correlation coeffieicent between jump sizes, which is a nuisance parameter.
    The LM test statistic for volatility jumps cannot be obtained in the case where jumps in returns and volatility are stochastically independent.
日時

4月22日(金 Friday) 16:50-18:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Lecture Hall No.3 on the 3rd floor of the New Economics Building

報告

Tatsuyoshi Okimoto (UCSD)

New Evidence of Asymmetric Dependence Structure in International Equity Markets: Further asymmetry in Bear Markets

Abstract A number of recent studies found two asymmetries in dependence structure in international equity markets: dependence tends to be high in (1) highly volatile markets and (2) bear markets. This paper further investigates the asymmetric dependence structure in international equity markets using two modern econometric techniques, namely, the Markov switching model and copula theory. While the former provides us natural and tractable models for processes with switching regimes, the latter can give us flexibility in describing asymmetric interdependence across observations. Combining these two theories, therefore, enables us to model regime switching dependence structure with sufficient flexibility. Using this flexible framework we indeed found there are two regimes: a high dependence regime with low and volatile returns, and a low dependence regime with high and stable returns. We also show that the bear regime is better described by the usage of an asymmetric copula with lower tail dependence, compared to using a normal copula. In addition, we show ignoring this further asymmetry in bear markets could be very costly for risk management.
日時

4月27日(水 Wednesday) 10:20-12:00

*COE Frontier Economics Lecture Series on Microeconometrics Methods

(April 27, May 11, 18, 25, June 1 開講予定)

場所

東京大学大学院経済学研究科棟 新棟3階 第2教室
at the Conference Hall No.2 on the 3rd floor of the New Economics Building

報告

雨宮健 Takeshi Amemiya (Stanford University)

Microeconometric Methods(Nested Logit Models, Tobit, Duration Models, and other current topics)

お知らせ

※4月27日(水)10:20-12:00 よりスタンフォード大学雨宮健 Takeshi Amamiya 教授による連続講義が開催されています。以降、5月11日、18日、25日6月1日の5回を予定しています。:Frontier Economics Lecture Series

連続講義に先立ち、4月13日(水)・20日(水)10:20-12:00に第203演習室において、国友直人教授により講義で前提とされる内容について予備的説明(Amemiya (1985)3 章・4章)が行われます。

日時

5月 6日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

楠岡成雄 Shigeo Kusuoka (東京大学数理科学研究科)

多期間リスク尺度と法則不変性(Multiperiod Risk Measures and Law Invariance)

要旨(Abstract)

近年、リスクの計量化が大きな関心を集め VaR, CVaR 等々のリスク尺度が
実務で用いられるようになった。 一方で、 CvaR 等にも問題があることが指摘され 期間構造を考慮したリスク尺度が研究されつつある。 講演では、最近の研究の紹介と同時に 森本祐二氏との共同研究の結果についても解説する。

日時

5月20日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

市村英彦 Hidehiko Ichimura (東京大学公共政策大学院)

Characterization of the Asymptotic Distribution of Semiparametric M-Estimators

(a joint work with Sokbae Lee, UCL)

要旨(Abstract)

A general formula for the asymptotic distribution of two-step semiparametric M-estimators are obtained. The first-stage nonparametric estimation can be profiled. We provide a simple formula for semiparametric M-estimators under regularity conditions that are relatively straightforward to verify. Calculating a formula for the asymptotic distribution involves Frechet differentiation of the expectation of an objective function. For many leading examples, this is often easy to derive. Our framework is illustrated by applying it to profiled estimation of single index quantile regression models.

日時

6月3日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

室町幸雄 Yukio Muromachi (ニッセイ基礎研究所)

モンテカルロ・シミュレーションと解析的手法を 融合したポートフォリオのリスク計測モデル  A New Portfolio Risk Measurement Method

要旨(Abstract)

ポートフォリオのリスク計測モデルの中核は将来価値(または潜在損失額)分布の推定法であり,その推定はモンテカルロシミュレーションを使う方法と解析的手法に大別される.前者では,モデルの柔軟な作り込みが可能であるが,ポートフォリオが大きくなると計算負荷が重くなりやすい。後者では,ポートフォリオが大きくなっても計算負荷は軽くて済むが,多くの近似が必要になるのでモデルの適切な適用範囲は狭い。本研究では,これら2つの手法を組み合わせて両者の長所を生かしたリスク計測手法(ハイブリッド法と呼ぶ)を提案する。実際にハイブリッド法で数値計算を行っところ,互いの長所が有効に機能する例もあれば,逆に短所が目立つ例もある.後者に対してはその解決案も提案する。実際のリスク計測では,ポートフォリオのリスク量を推定するだけでなく,ポートフォリオの各部分にどれだけのリスクが集中しているか(リスクの集中度)を算出することも重要である.数値例により,ハイブリッド法がリスクの集中度の推定に非常に有効であることも示す。

日時

6月10日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

北村祐一 Yuichi Kitamura (Department of Economics, Yale University)

Empirical Likelihood and Large Deviations

要旨(Abstract)

This talk explores connections between empirical likelihood and the theory of large deviations. The first half of the seminar is based on Kitamura (2001, Econometrica). In this paper the Hoeffding's (1965) results on optimal tests for multinomial models are extended to moment restriction models that are conventionally tested by Generalized Method of Moments (GMM). Empirical likelihood ratio tests are shown to be optimal in a Generalized Neyman-Pearson (GNP) sense, using a large deviation technique. Monte Carlo simulations suggest that empirical likelihood tests may also have better small sample properties than GMM. The second half of the seminar covers recent results obtained in Kitamura and Otsu (2004, mimeo.). It discusses asymptotically optimal estimation and parametric testing procedures for moment condition models using the theory of large deviations. First, it studies a moment condition model by treating it as a statistical experiment in Le Cam's sense, and investigates its large deviation properties. Second, it develops a new minimax estimator for the model by considering Bahadur's large deviation efficiency criterion. The estimator can be regarded as a robustified version of the conventional empirical likelihood estimator. Third, it considers a Chernoff-type risk for parametric testing in the model, which is concerned with the large deviation probabilities of type I errors and type II errors. It is shown that the empirical likelihood ratio test is asymptotically minimax in this context.

日時

7月1日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

星野伸明 Nobuaki Hoshino(金澤大学経済学部)

個票開示リスク評価とカウントデータモデリング

Evaluation of Disclosure Risk and Modelling Count Data

要旨(Abstract)

個票データセットを公開する前には、調査対象個体のプライバシーが暴露(開示)される可能性を正確に評価し、安全性を確認しなければならない。このような場合にプライバシーの危険性は、分割表各セルの度数の関数として扱うのが実際的である。ここで度数または「度数の度数」の同時分布をモデルとして扱いたい。このように考えると、離散多変数分布の新しい展開が見えてくる。本報告ではこの分野の問題意識を説明した後、非負整数上の無限分解可能分布を用いたモデルの構成方法を説明する。


日時

10月21日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

黒住英司 (一橋大学大学院経済学研究科)

Construction of stationarity tests with less size distortion

要旨(Abstract)

In this paper we propose a (trend) stationarity test that has good finite sample size even when a process is (trend) stationary with strong persistence. Our test may be seen as a modified version of the stationarity test proposed in Leybourne and McCabe (1994, LMC), but we use a different method to correct a series for serial correlation. Monte
Carlo simulations show that the empirical size of our test is close to the nominal one compared with the size of the original LMC test, while our test is more powerful than the LMC test with size-adjusted critical values. Our test is useful to distinguish between a (trend) stationary process with strong persistence and a unit root process.

日時

11月4日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

瀧本太郎 (九州大学)

A numerical evaluation method for factorizing the ARMA spectral matrix

要旨(Abstract)

Rational spectrum estimation based on a finite set of observations is a common practice in time-series analysis and it is usually conducted on the basis of time-domain ARMA representation of the data generating process.  But we would sometimes need direct factorization of a multivariate rational spectral density in case the density in question does not directly correspond to observed series but is a derived one from the spectrum of such series.  The paper presents a feasible computational procedure for canonical
factorization of rational spectral density matrix.  The method enables an stationary autoregressive-moving average representation of a process whose spectral density is given as a matrix of rational functions.

日時

11月18日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

渡辺重男(ニッセイ同和損害保険)

損害保険事業のリスクモデル

要旨(Abstract)

損害保険会社の経営環境は過去10年で激変し、それとともに損害保険会社のリスク管理に対する考え方にも大きな変化が見られる。中でも、会社全体のリスクを確率論的なモデルにより表現し、定量的に評価することで、資本の充分性の検証や資本効率の向上に資するといった考え方が急速に広まりつつある。本報告では、損害保険会社の事業リスクの中でも特に保険会社固有のリスクで ある保険引受リスクを中心に、リスクモデル構築の考え方や具体的な技法につ いて紹介する。

日時

12月5日(月 Monday) 10:30-12:00 ※曜日と場所に注意

場所

東京大学大学院経済学研究科棟 新棟12階 第1共同研究室

at the Conference Room No.1 on the 12th floor of the New Economics Building

報告

Herman K. van Dijk (Econometric Institute, Erasmus University Rotterdam)

Model uncertainty and model averaging in vector autoregressive processes

要旨(Abstract)

Economic policy decisions are often informed by empirical analysis based on accurate econometric modeling. However, a decision- maker is usually only interested in good estimates of outcomes, while an analyst must also be interested in estimating the model. Accurate inference on structural features of a model improves policy analysis as it improves estimation, inference and forecast efficiency. In this paper a Bayesian inferential procedure is presented which allows for unconditional inference on structural features of vector autoregressive (VAR) processes. We employ measures on manifolds in order to elicit uniform priors on subspaces defined by particular structural features of VARs. The features considered are cointegration, exogeneity, deterministic processes and overidentification.  Posterior probabilities of these features are used in a model averaging approach for forecasting and impulse response analysis. The methods are applied to three empirical economic issues: stability of Australian money demand, relative weights of permanent and transitory shocks in a US real business cycle model, and inflationary pressures due to an oil price shock in a UK structural VAR model.

日時

12月16日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

照井伸彦 Nobuhiko Terui (東北大学)

Discrete Choice Models based on Non-linear Stochastic Utility function and Their Applications

要旨(Abstract)

Extending standard linear stochastic utility function to non-linear function, we propose discrete choice models to accommodate non-linear response in panel data analysis. The hierarchical Bayes modeling with Markov chain Monte Carlo is applied to implement the models.   Then, two kinds of application are exposited. The first has theory-based and the second has empirical-based backgrounds for the modeling: 
1. Estimating Heterogeneous Price Thresholds
A brand choice model with heterogeneous price threshold parameters is used to investigate a three-regime piecewise-linear stochastic utility function. This study contributes to the modeling literature on discontinuous likelihoods in choice models. The empirical application using our scanner panel data set shows that the reference effect and loss aversion are more marked after price thresholds are taken into heterogeneous price response models. Furthermore, loss aversion is attenuated using price thresholds than by an aggregate (homogeneity) model without price thresholds. A customized pricing based on estimated price threshold is also discussed.
2. Modeling Heterogeneous Effective Advertising Stock Using Single- Source Data
This paper presents a model to test the hypothesis of advertising threshold effects and measure the effective advertising stock level using single-source data. The modeling includes the construction of advertising stock from exposure data having heterogeneous carryover parameter. We also gauge short-term and long-term effects of advertising at individual panels.

日時

12月21日(水 Wednesday) 4:50-6:10 ※曜日と場所に注意

場所

東京大学大学院経済学研究科棟 新棟12階 第2共同研究室

at the Conference Room No.2 on the 12th floor of the New Economics Building

報告

井上篤 Atsushi Inoue (North Carolina State University)

Efficient Estimation and Inference in Linear Pseudo-Panel Data Models

要旨(Abstract)

We consider pseudo-panel data models constructed from repeated cross sections in which the number of individuals per group is large relative to the number of groups and time periods. We show that (i) when group effects are neglected the OLS estimator does not converge in probability to a constant but rather to a random variable; and (ii) while the fixed-effects estimator is consistent, the usual t statistic is not asymptotically normal. We propose efficient GMM estimators and robust t statistics.

日時

1月6日(金 Friday) 4:50-6:10

場所

東京大学大学院経済学研究科棟 新棟3階 第3教室
at the Conference Hall No.3 on the 3rd floor of the New Economics Building

報告

西山慶彦(京都大学)

Statistical Theory of Rank Size Rule Rgression under Pareto Distribution

要旨(Abstract)

The rank-size rule has been examined by means of OLS estimation and the $t$ test in the literature. If we let $S_i$ be the $i$-th largest city in a country, it is often observed that $\log S_{(i)} \approx \alpha_0 + \alpha_1 \log i$ for some $\alpha_0>0$ and $\alpha_1<0$ and when $\alpha_1=-1$, this is defined as the rank-size rule.  However, since $S_{(i)}$ is heteroskedastic and autocorrelated, the $t $ statiistcs do not have a standard distribution. Indeed, we show $t \stackrel{p}{\rightarrow} \infty$ as the sample size increases even under the null hypothesis. The purpose of this paper is to obtain statistical properties of the OLS estimator of the rank size rule regression and the distribution of $t$ statistics under the Pareto distribution. We also propose more efficient estimation procedures and provide empirical applications of the theory for some countries.