ABSTRACT

We have assumed no distributional form for X. If X has a p-variate normal distribution with positive definite covariance matrix S, then the surfaces of constant

probability density are concentric ellipsoids and represents the major principal axis of these ellipsoids. In general under the assumption of normality of X, the principal components will represent a rotation of coordinate axes of its components to the principal axes of these ellipsoids. If there are multiple roots, the axes are not uniquely defined.