Guttman scaleの評価のための一研究

概要

論文の詳細を見る
This study is designed for the evaluation of Guttiman's scale from the practical point of view. A typical Guttman's and a non-Guttman's scale are constructed from the several items of Thurstone's Temperament Schedule which were administered to 855 male students of 4 colleges in Tokyo. Both scales are compared and evaluated in respects of unidimensionality, reliability, validity and practicability. The result of the study is summarized as follows: 1. We can conclude that Guttman's scale is superior to non-Guttman's scale in unidimensionality, reliability and internal validity. Both K-R and test-retest reliability of Guttman's scale are higher than those of non-Guttman's. In this experiment we see Guttman's scale is seven times as powerful as non-Guttman's in relation between reliability and the number of test items. 2. Correlation between G's scale and outside criterion is lower than that of non-G's scale, which may mean inadequacy of the experimental design. 3. G's scale is inadequate for the sharp discrimination of the subject group, especially extreme individuals in the group, because it is very difficult to construct G's scale which has a large number of items. 4. Construction of G's scale is more laborious than usual item analysis or G-P analysis, but less than factor analysis. 5. In order to construct G's scale easily, it seems to be wise to provide many question items inquiring overt and concrete behaviors of the subjects. 6. In order to get G's scale which has high sample-to-sample stauility, we need either decrease the number of items or in-crease the number of samples. At least 400 samples are needed to construct G's scale of four items. 7. G's scale, compared with usual scale, is not always inconvenient in practical use, i.e. test administration and scoring. Rather, it has the advantage of checking of applicability of the scale to a new subject group.
国際基督教大学の論文

著者

池田央
国際基督教大学

Guttman scaleの評価のための一研究

スポンサーリンク

概要

著者

関連論文

スポンサーリンク