Can Variances of Latent Variables be Scaled in Such a Way That They Correspond to Eigenvalues