(where g13, for example, is the return for the component #1 during month #3) and the returns of component "2" are g21, g22, ... g2M and so on and the deviations from the Mean (over all M months) can be displayed as an N x M matrix:
Further, we can display the fractions devoted to each asset via the 1xN matrix vector:
The M monthly deviations from the mean Portfolio Return, R = x1r1+x2r2+ ... +xNrN (as described in Part I) are the components of the 1xM matrix:
In order to compute the sum of the squares of this 1xM matrix, we multiply by its transpose. That is, if UT is a 1xM matrix, the sum of the squares of its components is the 1x1 scalar: UTU Hence, the sum of squares for our matrix (namely XT G) is the quadratic: Q = XT G { XT G)T = XT GGT X = XTWXwhere W is the covariance matrix. Its diagonal elements, when divided by N, give the square of the Standard Deviation - that is, the Variance - of each asset class. The off-diagonal elements give the co-variances between classes.
|