Quand le Professeur Chikio HAYASHI m'a proposé d'écrire un article pour la revue de la Société aux destinées de laquelle it préside, j'ai immédiatement compris que l'immense honneur qui m'était fait m'imposait d'accepter une telle offre. Mais grand a été mon embarras : les lecteurs de cette revue connaissent les travaux que le Professeur HAYASHI, a réalisés lui-même ou a dirigés: et je ne crois pas avoir rien à enseigner aux eleves d'un tel Maître ! J'ai donc pris le parti de traiter non des méthodes scientifiques que je cultive aprés lui, mais des espérances; qui n'appartiennent à personne, et auxquelles pourtant chacun confere des couleurs certes incommunicables, et cependant parfois dignes d'être reproduites par la peinture des mots, pour être exposées aux yeux de tous.
In the present paper, it is shown that the fluctuation of information takes approximately the same shape as that of word length, considered as cost of transmission in natural language, and that this tendency is remarkable at the main components of those two fluctuations. To confirm our statement, we used the spectral analysis for word length and for FB (Forward and Backward) information which reflects the interrelations among the words. From our experiment, we obtained a favorable result which suggests a explicit linear relation between those two fluctuations.
This paper describes an exploratory procedure for assessing the factor change, quasi three-mode factor analysis, which is applicable to longitudinal or multitraitmultimethod data. The model is a simplified case of Tucker's three-mode factor analysis, and formally equivalent to Kroonenberg and de Leeuw's Tucker 2 model. That is, it intends to decompose the standardized data matrix for each occasion into the product of three matrices, namely, the factor score matrix, the core matrix, and the factor loading matrix. It is illustrated that the patterns to which core matrices conform can distinguish the change of factor loadings from the change of factor scores. A practical least squares algorithm for the model is formulated, and a way of scalings of parameters which facilitates the interpretation of the results and the comparison with the output of other factor analytic methods is discussed. Analysis of a multitrait-multimethod matrix is attached as a numerical example.
A binomial error model which has a truncated beta distribution as its latent trait distribution is proposed for the analysis of test scores. This model can be used when the population of the examinees is assumed to be a restricted one. A procedure, the method of moments, is given to estimate the truncated points. An application of this model to the Japanese preliminary university entrance examination is also shown.
The empirical researches of Contingency Theory have given some prescriptive propositions available to organization design. But Contingency Theory has no mathematical models. In this paper, organization structure, management systems and efficiency of management systems are defined on the basis of the Modern Organization Theory, and models referred to by the decision makers in the organization are considered as models of the Sequential Decision Problems. The purpose of this paper is to show some normative propositions and Contingency Theoretic implications of them by using such concepts and models.