We propose a new technique of factor analysis with external criteria together with its computational procedure.
First of all, we shall discuss some relationships between the principle of analysis of variance and that of regression analysis by introducing the mathematical concept of
projection. Secondly, we shall show that two kinds of factors could be extracted, and that one of them would be closely related to the informations concerning the predetermined external criteria and another one would be independent of them. Thirdly, we shall show that our proposal might give a satisfactory answer to psychological interpretation of factors without assuming rotation of factors and the estimation of communalities. Finally, two numerical examples will be presented.
View full abstract