Consumer behavior has been analyzed from the viewpoint of competition between the store and the consumer. The present method analyzes the influence between a region with stores and its competing regions, and calculates the transition probability matrix based on the results of market research of consumer behavior. This study is suggested from the fundamental concept of the Input-Output Table for consumer behavior between regions, including stores.
Suppose that some data set consisting of several groups with variables of continuous type is given and discrimination of the data is required. Then two discrimination procedures are available. The first is canonical discriminant analysis(CDA)which uses the variables as they are, while the second is Hayashi's second method of quantification(Q2), canonical discriminant analysis for categorized data. This paper compares the two procedures from two points of view; the value of correlation ratio and the misclassification rate when the results of the analysis are applied to the original data themselves. As an illustration two data sets were used. The first is based on medical claim units for the elderly compiled by Kosyu Eisei Shinkokai. The second is the famous iris data treated often in discriminant analysis. Neither of the two procedures can be said better than the other.
The purpose of this study was to develop a scale for measuring cognitive stress in order to assess mental health education program for high school students. First, the content of stressor was collected using the open-ended questionaaire from 1, 196 Korean high school students in Seoul, and 149 items were obtained. Next, stress magnitude of each stressor was appraised with 4-point rating scale by1, 188 Korean high school students in Seoul. On the basis of the results of principal component analysis and item analysis, subscales concerning daily hassles, e. g., home life, school life, future, friendship, social support, health and so on, were constructed. Eighteen items concerning life event were adopted as one subscale, which can be calculated in terms of 4-point rating scale. As a result, 36subscales consisted of 98 items were chosen. Moreover, 14integrated scales were constructed by a principal component analysis of these subscales. Lastly, percentile and T-score were prepared as a norm.
SDS(Self-rating Depression Scale)developed by Zung in 1965 is widely applied to psychiatric studies. The existence of the response bias in SDS, however, has been suggested by several researchers. The purpose of the present study was to examine precisely the response bias in the Japanese version of the SDS. Data from 2, 097 Japanese high school students are shown to illustrate the response bias related to the positive or negative expressions of items in the scale with statistical methods of analysis, together with several sets of empirical data from otherstudies. The findings were discussed in terms of statistical view. Apart from the main subjects, we point out some incorrectness of Zung(1967b)in Appendix.
Recent vast development of high-speed computer and large storing facilities together with network environment enables us to construct very large or sometimes huge databases such as giga-byte level. In order to extract novel, useful and also interpretable information from those databases, new technology becomes potentially important and being called for in various research and commercial fields. Data mining and knowledge discovery in databases(KDD)are the names given to such activity which involves database technology, machine learning, data visualization and statistics. Since we deal with data, statistics and statisticians are to be expected to play an essential role in data mining and KDD as well. Data mining, however, differs from traditional statistics on some dimensions, in which?gscale?his the most important one. The present paper first reviews recent achievement of data mining and KDD from a statistical viewpoint. Although some differences between data mining and traditional statistics are pointed out, it will be emphasized here that those two areas are closely related and should be recognized as complement to each other. Some important research areas are also discussed. One important message of this review is that statisticians should be involved in this new activity and should play a vital role in developing new methodologies and also in finding various application areas in practice.