抄録
Kaiser's normalization is widely used in factor rotation. In this paper the asymptotic standard errors for rotated parameters are obtained when Kaiser's normalization is employed for the direct oblimin method which is one of the most frequently used oblique rotations. The method of estimating the standard errors is based on the augmented information matrix for parameters with restrictions. A Monte Carlo simulation is carried out to confirm the accuracy of the method. Further, it is shown by artificial data that the values of the standard errors with Kaiser's normalization can be significantly different from those without the normalization. That is, Kaiser's normalization tends to decrease the standard errors of the loadings for the variables with small communalities and to increase those of the correlations among oblique factors.
