JOURNAL OF THE JAPAN STATISTICAL SOCIETY
Online ISSN : 1348-6365
Print ISSN : 1882-2754
ISSN-L : 1348-6365
Volume 30, Issue 1
Displaying 1-9 of 9 articles from this issue
  • Eiji Nakashima
    2000 Volume 30 Issue 1 Pages 1-15
    Published: 2000
    Released on J-STAGE: April 14, 2008
    JOURNAL FREE ACCESS
    A model for the analysis of repeated ordered polytomous data using the generalized estimating equations(GEE)method is presented. Our model is a direct extension of McCullagh's model(1980)for longitudinal data. For independence estimating equations, this model is equivalent to the models proposed by Miller et al.(1993)and Lipsitz et al.(1994). When one extends this model to allow for correlated observation, our model is not equivalent to theirs. Furthermore, our model can estimate the correlations using a simpler working covariance matrix in the second set of estimating equations than that of Miller et al., and it allows for the use of the AR(1)correlation structure, which is not possible using either of the others. Two examples of the analyses of longitudinal ordered polytomous data are given, one for judgmental ordinal data and the other for grouped continuous ordinal data. Among several working correlation models, a relatively simple working correlation structure suffices for each data set.
    Download PDF (1263K)
  • Etsuo Miyaoka, Miki Kajikawa, Bruce Smith
    2000 Volume 30 Issue 1 Pages 17-25
    Published: 2000
    Released on J-STAGE: April 14, 2008
    JOURNAL FREE ACCESS
    This paper presents a new model for the analysis of binary data from cross-over trials. The model is parameterized by the conditional probabilities of period two responses given period one outcomes and the period one outcome probabilities. Parameter estimation and hypothesis testing are carried out using standard procedures for generalized linear models. An analysis of drug efficacy is presented to illustrate the methods, and a simulation study is used to assess the small sample behavior of the method.
    Download PDF (778K)
  • Makoto Aoshima, Yoshikazu Takada
    2000 Volume 30 Issue 1 Pages 27-41
    Published: 2000
    Released on J-STAGE: April 14, 2008
    JOURNAL FREE ACCESS
    The problem of constructing a set of fixed-width simultaneous confidence intervals for the treatment-control differences of means is considered for several independent normal populations with a common unknown variance. A two-stage procedure is developed for such inference and its asymptotic characteristics are studied up to the second order. The associated results are also provided for all pairwise comparisons problem. Finally, performances of the proposed two-stage procedure are compared for both small and moderate sample sizes in several cases.
    Download PDF (1044K)
  • Hidehiko Kamiya, Akimichi Takemura
    2000 Volume 30 Issue 1 Pages 43-51
    Published: 2000
    Released on J-STAGE: April 14, 2008
    JOURNAL FREE ACCESS
    We consider the same problem as in Kamiya and Takemura(1997), but for discriminant analysis on(n-1)-dimensional unit sphere Sn-1. That is, we regard pairwise discriminant analysis of m populations on Sn-1 as a process to generate rankings among the populations, and give a formula for the number of generated rankings.
    Download PDF (733K)
  • Wataru Yamamoto, Nobuo Shinozaki
    2000 Volume 30 Issue 1 Pages 53-63
    Published: 2000
    Released on J-STAGE: April 14, 2008
    JOURNAL FREE ACCESS
    This article investigates two principal points of location mixtures of spherically symmetric distributions. We give a lemma which enables us to restrict the region to search principal points, and, with this lemma, prove a subspace theorem which states that there exist two principal points in the linear subspace spanned by the component means. We also give a sufficient condition for uniqueness of two principal points for two component cases. These results can be applied to a class of spherically symmetric distributions, which includes multivariate normal and t distributions.
    Download PDF (892K)
  • Yushan Xiao
    2000 Volume 30 Issue 1 Pages 65-73
    Published: 2000
    Released on J-STAGE: April 14, 2008
    JOURNAL FREE ACCESS
    This paper considers empirical Bayes estimators from the unbiased prediction point of view. The consideration is devoted to the normal case. It is shown that some empirical Bayes estimators are the best or the best conditional unbiased.
    Download PDF (691K)
  • Norihisa Tsuga, Masanobu Taniguchi, Madan L. Puri
    2000 Volume 30 Issue 1 Pages 75-87
    Published: 2000
    Released on J-STAGE: April 14, 2008
    JOURNAL FREE ACCESS
    Suppose that we observe(xt, yt)from the errors-in-variables model : xttt, yt=βξtt, where{δt}and{εt}are i.i.d.measurement errors. Here we assume that{ξt}is a non-Gaussian stationary process with zero mean and spectral density fξ(λ). For this model, some estimators for β have been proposed in the literature. However, they are constructed under the assumption that the data are independent normal variates. Thus they do not contain the dependent structure of the data(e.g., time-lagged sample covariances, etc.). In this paper we propose a new class Λ of estimators of β, which is defined under consideration for dependent structures of (xt, yt, ξt). Then the asymptotic distribution of β^^^∈Λ is derived. We give an asymptotically optimal estimator in this class. Comparison with the existing estimators is also discussed. Since the asymptotic variance of β^^^ is complicated we have illuminated some aspect of the asymptotics numerically using Mathematica.
    Download PDF (836K)
  • Takakatsu Inoue
    2000 Volume 30 Issue 1 Pages 89-104
    Published: 2000
    Released on J-STAGE: April 14, 2008
    JOURNAL FREE ACCESS
    The double f-class generalized ridge regression estimator is known to be a modified estimator of the generalized ridge regression estimator. In this paper, the exact distribution and density functions of the 'expanded'double f-class generalized ridge estimator are derived. Based on the density function derived, the performance of the relative efficiency of the expanded estimator to the ordinary least squared estimator is also evaluated by numerical analysis for two prior distributions
    Download PDF (1113K)
  • Jong-Wuu Wu
    2000 Volume 30 Issue 1 Pages 105-113
    Published: 2000
    Released on J-STAGE: April 14, 2008
    JOURNAL FREE ACCESS
    In the present paper, the finite mixture of exponential distributions(Everitt and Hand(1981)and Johnson et al.(1994))is characterized by using the recurrence relations of conditional moments of nonadjacent upper record values. Furthermore, we also mention a similar theorem to characterize the finite mixture of exponential distributions by using the recurrence relations of conditional moments of nonadjacent lower record values.
    Download PDF (644K)
feedback
Top