Let's assume we have 4 independent random vectors that have values $\in [0.0, 1.0]$ and a lot of elements. Label them: $\{r_1,r_2,r_3,r_4\}$. Now, let's consider a simple case of 3 variables and try understanding the relation between them. Consider two distinct sets:

$S_1 = \{ r_1 + r_2 , r_1 + r_3, r_1 + r_4\}$


$S_1 = \{ r_1 + r_2 , r_2 + r_3, r_3 + r_1\}$

Both sets have identical covariance matrices because each pair has only one common random vector:

$ Cov(S_1) = Cov(S_2) = \left( \begin{array}{ccc} 1.0 & 0.5 & 0.5 \\ 0.5 & 1.0 & 0.5 \\ 0.5 & 0.5 & 1.0 \\ \end{array} \right) $

but obviously the relationship between the vectors in a set are very different. How to quantify and estimate the relationship between all the vectors in a set?

PS: this question if a follow up from my previous question on covariance matrices.

  • $\begingroup$ What does it mean for a vector to have values in $[0.0,1.0]$? What does it mean for a vector (or something else?) to have "a lot of elements"? $\endgroup$ Feb 15, 2016 at 23:34
  • $\begingroup$ Hey @DilipSarwate, i meant that random vectors $r_i$ have some arbitrary large number of elements (dimensions) such that the errors on our $Cov$ estimates are small. By range $[0,1]$, I mean that all the elements within the vector are within that range. I don't think it changes anything, but it is easier for me to understand if I think about specific example. thanks! $\endgroup$
    – Karolis
    Feb 15, 2016 at 23:53


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.