It is the purpose of the data analysis to clarify the characteristic of the phenomena and to find the meaningfull relations among them. From the point of view, a graphical description of the data is effective for the understanding of the structure of them and, in paticular, is important for the multiple analysis. In this paper, we consider the method of the graphical description of the correlation coeffcient, the multi correlation coefficient and the partial correlation coeffcient in the multiple analysis.It makes useful for the visualization of the complex relationship among the data. Moreover it applies to the choice of the variables and the multicollinearity.
SLTA(Standard Language Test of Aphasia)was administered to a total of 313 aphasic subjects and the test records were analyzed by means of factor analysis and scalogram analysis. Factor analysis gave a simple structure which allowed a classification of test variables into three groups representing three language modalities; group A of writing, group B of speech and group C of comprehension. Separate scalogram analyses of these groups A, B and C revealed an almost ideal unidimensionality for scaling in respective groups. Distribution of the subjects was displayed by three axes of comprehension, speech and writing corresponding to three edges of a cuboid. The subjects were laid along the three edges, jointing each other to form a“gzigzag way”in the cuboid. Thus an ordering of comprehension, speech and writing, from severe impairment to mild, was found along the“zigzag way”which represented a continuous3-dimension structure.
An equating method of two scales based upon two separate tests using common subjects was proposed in the previous paper(Noguchi, 1983). In that method, each subject's scale values are estimated on the two separate scales, and then the principal component for the whole group of examinees is extracted, which corresponds to the equating equation. In this paper, using computer simulation, the validity of Noguchi's method is examined in various situations, changing the difficulty levels of the two tests, the distribution of subjects and the number of subjects. As a consequence, Noguchi's equating method proved to be valid in situations where the two tests are not so different in difficulty levels and the subjects'ability distributes for a wide range.
The two-way table, whose element eij represents the wholesale dependence ratio of each prefectures i to the three dominant prefectures j(Tokyo, Aichi, Osaka), is considered. A mapped representation of the dominant-subdominant relation in this table to the two dimensional Euclid space is shown, as the method of perceiving the structure of this relation macroscopically, and detecting the prefectures behaving contrary to the whole structure. The procedure is devised so that the coordinate of the prefecture i is firstly arranged to the natural region of this map corresponding to the order of the magnitudes of eij's for each j, then the distance-type objective function similar to the Kruskal's stress is minimized.
In this paper, we tried to compare the results of the attitude surveys toward nature in several cities using several techniques of multidimensional analysis. The surveys were carried out in the following cities by Research Group of Forest-Environmental Problems and etc. Japan: Asahikawa, Tsuruoka, Kushibiki, Ina, Miyazaki and Tokyo Germany: Freiburg i. Br., Neuenburg, Gottingen and Hannover Austria: Wien, Linz and Salzburg France: Nancy The questionnaire are classified in the following types;(1)The affection of forests in daily life.(2)The mystic and holy feeling about trees and forests.(3)The propriety of forest management.(4)The preference of the forest type by the paired comparison method using the five suits of photographs. We can say that the results of our surveys make clear a part of the difference of the attitudes toward forest of‹Tokyo›, ‹Japan(excluding Tokyo)›, ‹Germany and Austria›, ‹Nancy›.