Journal of the Japan Statistical Society, Japanese Issue
Online ISSN : 2189-1478
Print ISSN : 0389-5602
ISSN-L : 0389-5602
Search
Advanced search
OR
Browse
Search
Volume 45, Issue 1
Displaying 1-13 of 13 articles from this issue
Show all abstracts
    • |<
    • <
    • 1
    • >
    • >|
Articles
  • Discrete Bartlett-Type Transformed Statistics Using a Discontinuous Term of Asymptotic Expansion
    Yuri Sekiya, Nobuhiro Taneichi
    2015 Volume 45 Issue 1 Pages 1-17
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.1
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    In the multinomial goodness-of-fit test, it is well known that the log-likelihood ratio statistic T has a limiting chi-square distribution under a simple null hypothesis. Siotani and Fujikoshi (1984) derived an asymptotic expansion for the null distribution of the log-likelihood ratio statistic T under a simple null hypothesis. They proposed J1+Ĵ2 as an approximation of the lower probability of T, where J1 is a term of multivariate Edgeworth expansion and Ĵ2 is a first order approximation of a discontinuous term. In this paper, we introduce discrete Bartlett-type transformed statistics T* and T** of the log-likelihood ratio statistic T that are constructed from J1 and Ĵ2. By numerical comparison, we found that the accuracy of approximation of the log-likelihood ratio statistic T to a chi-square distribution is improved by T* or T**.
    View full abstract
    Download PDF (688K)
  • The sp-transform of Probability Distributions
    Hiroyuki Takeuchi
    2015 Volume 45 Issue 1 Pages 19-40
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.19
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    Takeuchi (2013) showed that the saddlepoint uniquely determines its distribution under the existence of the analytic characteristic functions. In this paper we shall express the saddlepoint as an envelope (sp-curve) on the statistical model manifold ℳ of the normal distributions. The length of the sp-curve on ℳ is shown to be not depending on the coordinates,　and is crucial for the asymptotic normality of the standardized sample mean. The sp-transform is defined to get the sp-curves, and is shown to be bijective. With using this property, we can say that any distribution which has saddlepoint can be generated by a family of the normal distributions.
    View full abstract
    Download PDF (1578K)
Special Section: Econometric Analysis of Spatial Data
  • Estimation of Marginal Effects and Their Spillover Effects in Productivity of Prefectures
    Yoshihiro Ohtsuka
    2015 Volume 45 Issue 1 Pages 41-57
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.41
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    This study estimates the spillover effect of productivity function, and measures marginal effects of the production on the capital and labor force, using the spatial dynamic panel data model with prefecture panel data from 2001 to 2009. We introduce the spatial dynamic panel data model with random effects and the Bayesian methodology, and explain to decompose the marginal effects into the direct effect and indirect effect in the short term and the long term. From the empirical results by using Markov chain Monte Carlo (MCMC) method, it is found that the prefectural productivity depends on spatially and serial correlations. Futhermore, the indirect effect of the labor force is effective in the short term. On the contrary, the direct effect of the labor force in the long term is larger than that in the short term.
    View full abstract
    Download PDF (766K)
  • The Measurement of the Direct and Indirect Effects in Road Use
    Kazuhiko Kakamu
    2015 Volume 45 Issue 1 Pages 59-68
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.59
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    In the linear regression model, as the marginal effect with respect to the independent variable is the estimated parameter itself, it is intuitive and tractable. However, in the spatial econometric model, as the marginal effect with respect to the independent variable depends on the other regions' variables, it is difficult to facilitate it. In this paper, we explain the method proposed by LeSage and Pace (2009) to decompose the marginal effect into direct effect, which arises from the own region, and indirect effect, which arises from the other regions, and introduce the application to the road data in Japan.
    View full abstract
    Download PDF (704K)
  • Migration Response to High Unemployment Rates: Spatial Econometric Analysis Using Japanese Municipal Data
    Keisuke Kondo
    2015 Volume 45 Issue 1 Pages 69-98
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.69
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    Using the Japanese municipal data-set between 1980 and 2010, this paper empirically analyzes the underlying mechanism of interregional labor migration. Since the existing literature shows that regional disparities in unemployment rates have decreased gradually, our main concern is to uncover how the labor migration contributes to reducing the unemployment disparities. A novel approach to the migration analysis is to incorporate interregional dependency in migration decisions by using spatial econometric models. Our estimation results show that high unemployment rates play a role of push factor for migration and that there is a statistically significant and positive spatial dependence in interregional migration. Furthermore, we find a negative relationship between percentage changes in relative unemployment rates and out-migration rates. Therefore, our results suggest that (in)sufficient out-migration of the unemployed from high (low) unemployment regions leads to low (high) regional unemployment rates in subsequent periods.
    View full abstract
    Download PDF (8458K)
  • Spatial Distribution of China's Manufacturing Firm-Level Productivity: An Attempt at Blockwise GWR
    Kuang-hui Chen, Yoshihiro Hashiguchi
    2015 Volume 45 Issue 1 Pages 99-118
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.99
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    This paper visualizes the spatial distributions of China's state and non-state manufacturing firms' productivity. It uses a large firm-level data set and proposes methods of blockwise GWR (Geographically Weighted Regression) and mixed GWR, methods which allow us to estimate spatially varying parameters with large micro data. They enable flexible model building and can be applied in a wide range of statistical analyses. Using a blockwise mixed GWR model that has fixed coefficient variables to control for firm heterogeneity, a large regional disparity in manufacturing productivity is clarified in China. Productivity growth has been high in the area south to the Yangtze River (particularly high in areas around Changsha City), and the productivity disparity has been large in the North and the Northwest regions.
    View full abstract
    Download PDF (11176K)
Special Section: Selected Topics in Bayesian Inference
  • Duality Between Likelihood Maximization and Shannon Entropy Maximization in Bayesian Prediction
    Toshio Ohnishi
    2015 Volume 45 Issue 1 Pages 119-141
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.119
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    We formulate a risk minimization problem in Bayesian prediction in the framework of Bayesian model averaging. Goodness of prediction is evaluated by adopting the dual Kullback-Leibler divergence losses. Duality between likelihood maximization and Shannon entropy maximization is revealed through that of the divervence losses.
    View full abstract
    Download PDF (705K)
  • A Review: The Stein Paradox and Bayes Theory
    Yuzo Maruyama
    2015 Volume 45 Issue 1 Pages 143-170
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.143
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    The Stein paradox on estimation and prediciton under multivariate normal distribution and spherically symmetric distribution is reviewed. In particular, estimation of a multivariate normal mean with known variance, the relationship of the esimation of mean vector and predictive density and parameter estimation of linear regression model are treated. Bayesian procedures play an important role as a remedy for the Stein paradox.
    View full abstract
    Download PDF (775K)
  • Bayesian Inference Under a Family of Prior Densities Containing Various Amounts of Information
    Takemi Yanagimoto
    2015 Volume 45 Issue 1 Pages 171-191
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.171
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    Bayesian inference under a weakly informative prior density is a key research subject for pursuing the construction of widely applicable Bayesian procedures under various practical situations. A non-informative prior density can be defined as a limit of a sequence of weakly informative prior densities, and reversely such a sequence can connect a proper prior density with a weakly informative prior density. The behavior of a Bayesian procedure under a weakly informative prior density is an important issue, though the current theory of the empirical Bayes method often treats it as a least favorable case. This article reviews inferential procedures under the assumption of a family of prior densities connecting from a degenerated density on a known point to a non-informative density. Then, we emphasize the need for introducing suitable definitions of Bayesian likelihoods. Lindley's paradox and the actual difficulty in defining a non-informative prior density are referred to as two attractive implications.
    View full abstract
    Download PDF (780K)
Special Topic: The JSS Prize Lecture
  • Numerical Computation for the Eigenvalue Distributions of a Wishart Matrix
    Takakazu Sugiyama, Hiroki Hashiguchi
    2015 Volume 45 Issue 1 Pages 193-210
    Published: September 30, 2015
    Released on J-STAGE: May 30, 2016
    DOIhttps://doi.org/10.11329/jjssj.45.193
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    Wishart matrix is one of typical random matrices and a constant times of a sample covariance matrix. The larger eigenvalues and corresponding eigenvectors of the sample covariance matrix are important to assess results from a sample in multivariate statistical analysis. On the other hand, the smaller ones are related to collinearity in a regression model. This paper discusses numerical computation for the istributions of the eigenvalues and the largest eigenvector for a Wishart matrix, and also show that approximations based on normal and chi-square distributions have high accuracy.
    View full abstract
    Download PDF (786K)
Book reviews
    • |<
    • <
    • 1
    • >
    • >|
feedback
 Top