Ouyou toukeigaku
Online ISSN : 1883-8081
Print ISSN : 0285-0370
ISSN-L : 0285-0370
Search
Advanced search
OR
Browse
Search
Volume 30, Issue 3
Displaying 1-3 of 3 articles from this issue
Show all abstracts
    • |<
    • <
    • 1
    • >
    • >|
  • A Diagnostic Method for Incomplete Data Based on the Distance between Data Patterns
    Yoshiyuki Inaba
    2001 Volume 30 Issue 3 Pages 131-143
    Published: March 30, 2002
    Released on J-STAGE: June 12, 2009
    DOIhttps://doi.org/10.5023/jappstat.30.131
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    We could not handle social survey data without nonresponse. General approaches to detect the conditions of incomplete data are based on variables. In this paper, we propose a diagnostic method for incomplete data based on data patterns. Proposed method is applicable to mixed categorical and continuous variables. The results of diagnostic are helpful for applying the pattern-mixture model to incomplete data. We indicate that the method can be useful by using examples from a social survey dataset.
    View full abstract
    Download PDF (641K)
  • Nonparametric Density Estimate by Piecewise Linear Kernel Function
    Keiko Yamamoto, Masahiko Sagae
    2001 Volume 30 Issue 3 Pages 145-160
    Published: March 30, 2002
    Released on J-STAGE: June 12, 2009
    DOIhttps://doi.org/10.5023/jappstat.30.145
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    Kernel density estimation is one of the well-known technique in nonparametric density estima-tions. Kernel method is constructed by placing a kernel function disposed at each data point and it can obtain smooth estimations. We note that this method has cost of calculation increasing in proportion with the number of data. Its computational burden is relieved by the development of the computer, however, if we choose a continuous probability density function, then the calculation in-creases more. Therefore we suggest the trapezoid kernel function generated by the piecewise linear function, because we can reduce the cost of calculating the estimate by using trapezoid kernel. We show that it is easy to compute and has the better efficiency compared with other kernel functions. We discuss about the property of trapezoid kernel function focused on the calculational efficiency and show the numerical example, also.
    View full abstract
    Download PDF (570K)
  • A Study on Treatments of Ties in Ward's Clustering Method
    Koichi Toita
    2001 Volume 30 Issue 3 Pages 161-184
    Published: March 30, 2002
    Released on J-STAGE: June 12, 2009
    DOIhttps://doi.org/10.5023/jappstat.30.161
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    Ward's method or Ward's error sum of squares (ESS) method is a popular procedure in hier-archical cluster analysis. The results of Ward's method depend on indexes of objects, especially when there are ties in dissimilarities. The paper studies treatments of ties and proposes an extended formula to simplify the calculations of ESS.
    View full abstract
    Download PDF (855K)
    • |<
    • <
    • 1
    • >
    • >|
feedback
 Top