Journal of the Japanese Society of Computational Statistics

CIRCLE STRUCTURE DERIVED FROM DECOMPOSITION OF ASYMMETRIC DATA MATRIX

Takayuki Saito

2002 Volume 15 Issue 1 Pages 1-18
Published: 2002
Released on J-STAGE: December 09, 2009

DOIhttps://doi.org/10.5183/jjscs1988.15.1

JOURNAL FREE ACCESS

Show abstractHide abstract

Among many approaches to analysis of asymmetric data, this paper is concerned with the matrix decomposition approach. Focussing on Gower's procedure, we study conditions to derive circle structure from an asymmetric data matrix. Results are presented in terms of some theorems and lemmas, followed by discussion with numerical illustrations.

View full abstract

Download PDF (1441K)
DETERMINATION OF SAMPLE SIZES FOR USE IN CONSTRUCTING CONFIDENCE INTERVALS FOR A BINOMIAL PARAMETER

Manabu Iwasaki, Noriko Hidaka

2002 Volume 15 Issue 1 Pages 19-32
Published: 2002
Released on J-STAGE: December 09, 2009

DOIhttps://doi.org/10.5183/jjscs1988.15.19

JOURNAL FREE ACCESS

Show abstractHide abstract

This article considers exact and approximate confidence intervals for a binomial parameter p. Specifically, we will deal with the problem of determination of sample sizes which guarantee the probability that 100 (1-α) % confidence intervals for p do not include pre-specified constants is greater than 1-β. It is shown here that the coverage probability of confidence intervals is not an increasing function of the sample size, which is a consequence of the discreteness of binomial distributions. Illustrative numerical examples and some theoretical consideration for such anomalous behavior are given. We also briefly treat the case that the initial guess of the true binomial parameter is vague.

View full abstract

Download PDF (1293K)
STATISTICAL INFERENCES OF GROUPED DATA COMBINED WITH UNGROUPED DATA FROM THE POWER-NORMAL DISTRIBUTION

Toshimitsu Hamasaki, Masashi Goto

2002 Volume 15 Issue 1 Pages 33-52
Published: 2002
Released on J-STAGE: December 09, 2009

DOIhttps://doi.org/10.5183/jjscs1988.15.33

JOURNAL FREE ACCESS

Show abstractHide abstract

We describe an exploratory approach to analyzing grouped data in which an underlying distribution of data is unknown or there is no strong knowledge of the process of generating the data. In addition to the most elementary situation where one variable of interest and its observations are given in a grouped form such as a frequency table, we consider the two types of mixed grouped data, i.e., ungrouped data available in the tail and right and/or left censored data. We develop the procedure of the parameter estimation and investigate the asymptotic properties of the estimates for each type of grouped data. We also provide two examples to illustrate the proposed approach. The results show that the proposed approach would help to “regularize” the data even when it does not yield normality.

View full abstract

Download PDF (1738K)
LIKELIHOOD RATIO TEST FOR MEAN VECTOR WITH SPECIFIED DIRECTION BASED ON CONDITIONAL DISTRIBUTION

Tsunehisa Imada, Hideyuki Douke

2002 Volume 15 Issue 1 Pages 53-64
Published: 2002
Released on J-STAGE: December 09, 2009

DOIhttps://doi.org/10.5183/jjscs1988.15.53

JOURNAL FREE ACCESS

Show abstractHide abstract

The aim of this study is to construct simple methods of a multivariate one-sided test and a multivariate two-sided test when the covariance matrix is unknown. First we assume that the mean vector has a specified direction as Tang, Gnecco and Geller (1989) considered. Then we set up two kinds of composite hypotheses and derive the likelihood ratio test statistic for our composite hypotheses. However the explicit formula of distribution of this statistic does not exist. Thus we use the conditional distribution considered by Wang and McDermott (1998) to determine the critical value independently of the unknown covariance matrix. Finally we compare our methods with the method of Wang and McDermott (1998) in terms of a numerical example regarding the power of the test.

View full abstract

Download PDF (959K)
DISTRIBUTED PROCESSING FUNCTIONS OF A TIME SERIES ANALYSIS SYSTEM

Yoshikazu Yamamoto, Junji Nakano

2002 Volume 15 Issue 1 Pages 65-77
Published: 2002
Released on J-STAGE: December 09, 2009

DOIhttps://doi.org/10.5183/jjscs1988.15.65

JOURNAL FREE ACCESS

Show abstractHide abstract

In this paper, we propose the use of the server/client system in statistical analysis. We show that the interactive response speed of a time series analysis system TISAS (TIme Series Analysis Supporting system) has been improved by adding functions for distributed processing by using the Parallel Virtual Machine (PVM) library. TISAS was originally designed for building an object-based graphical user interface (GUI). The Tcl/Tk interpreter language was chosen as the main programming language, because it was designed especially for building GUI applications easily. We reconstruct TISAS by adopting a server/client system structure without changing the style of the GUI by using TkPVM, which is an implementation of the PVM library through Tcl/Tk.

View full abstract

Download PDF (11612K)
List of authors and titles : The 16th

2002 Volume 15 Issue 1 Pages 79-84
Published: 2002
Released on J-STAGE: December 09, 2009

DOIhttps://doi.org/10.5183/jjscs1988.15.79

JOURNAL FREE ACCESS

Download PDF (396K)

Register with J-STAGE for free!