Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I understand that the outer product of two vectors, say representing two detrended time series, can represent a cross-correlation (well covariance) matrix.

I also know that the inverse of a correlation matrix represents the partial correlations between two variables. Geometrically, I know that the partial correlation between two variables is the angle formed by the projection of their residuals when regressed against all other variables onto the surface perpendicular to all other variables.

I'm wondering how these two relate. That is, I know the interpretation of the inverse of matrix (partial correlation) but not the matrix or its construction.

share|improve this question

closed as not a real question by whuber Sep 29 '12 at 21:44

It's difficult to tell what is being asked here. This question is ambiguous, vague, incomplete, overly broad, or rhetorical and cannot be reasonably answered in its current form. For help clarifying this question so that it can be reopened, visit the help center.If this question can be reworded to fit the rules in the help center, please edit the question.

Your reference to "outer product" does not accord with my understanding of this operation, which I believe is a conventional one. The rank of any outer product is at most one, which would produce a highly degenerate matrix: that's not what one expects of a covariance. Could you indicate what your "outer product" operation is? – whuber Sep 14 '12 at 22:27
am sure he refers to the sum of all the outer products of $x_i$ with itself over $i$. That is just the sum of cross-products ... – kjetil b halvorsen Sep 14 '12 at 23:09
By outer product, I mean $ \mathbf{x}\mathbf{y}^T$. I thought that for two random vectors, $\mathbf{x}$ and $\mathbf{y}$ the covariance matrix would be $ E\left(\mathbf{xy^T}\right) = E\left(\mathbf{x}\right) E\left(\mathbf{y}\right)^T $. – mac389 Sep 15 '12 at 2:58
Maybe I don't understand. Suppose that the random vectors $X$ and $Y$ have mean vectors $u=\mathrm{E}[X]$ and $v=\mathrm{E}[Y]$, respectively. The matrix $A=u\,v^T$ can't be, in general, a covariance matrix, because, suppose that $X$ and $Y$ are such that the inner product $\langle u,v\rangle<0$. Then, we have $\langle u,Au\rangle=u^TAu=u^Tu\,v^Tu=||u||^2 \langle v,u\rangle<0$, and $A$ is not positive definite. – Zen Sep 16 '12 at 2:23