Bulletin of the Computational Statistics of Japan

In table tennis, the number of hits---representing the number of rallies---is used as a quantity that represents the characteristics of players or plays. This study proposes the zero-distorted generalized cross geometric distribution and the double zero-distorted generalized cross geometric distribution as models that can take into account the difference in the return rate between servers and receivers, and can flexibly express zero inflation/deflation. For these probability models, we derive parameter estimation methods by the maximum likelihood method and approximate confidence intervals for the parameters. Furthermore, we apply them to actual data from the T-League of Japan and simulate the coverage probability of the approximate confidence intervals.

View full abstract

In this study, we propose a visualization method for market segmentation using a supervised self-organizing map. The label of the action selection segment provided by the data provider Video Research Inc. was used as a teacher signal. We extracted feature data by converting the data on questionnaires and TV viewing behaviors using non-negative matrix factorization. The proposed method realized the visualization of the relationship between market segments and the factors affecting them. Consequently, the presentation of advertisements and recommendations and the acquisition of knowledge for the development of new products can become more efficient.

View full abstract

In this paper, we make two suggestions for a program to improve the real-time viewing rate on a television talk show of the golden time. The provided data was aggregated by statistical analysis software, we performed a multivariate analysis such as factor analysis or cluster analysis on the data. And we clarified the feature of a program with a high real-time viewing rate of talk show programs of the golden time. From the results of the analysis, we propose a television program based on its features, to promote real-time viewing and ratings.

View full abstract

Show abstractHide abstract

Graph representation makes it possible to intuitively grasp complex data structures and detailed features, and to easily interpret data features that cannot be grasped by numerical values alone. In this paper, we focused on the constellation graph and the expanded constellation graph that applies to the constellation graph. The constellation graph can grasp the mean value and the degree of variation at a glance, and the expanded constellation graph can easily grasp the characteristics of variables in each individual. In order to build an environment where constellation graphs can be easily executed on a web basis, we developed software using an R shiny package, and introduced how to use it with application examples. By using shiny, it is possible to develop interactive web applications. So, we were able to build an environment in which constellation graphs can be flexibly used for various data while performing interactive operations.

View full abstract

Download PDF (4139K)

Register with J-STAGE for free!