Suppose we have a matrix $A_n = \frac{1}{n}\sum_{i=1}^nX_i X_i^T$, where $X_i$ is a $p$-dimensional random-vector. We also know that $E(XX^T) = \Sigma_{p \times p}$. Let us denote the $j$-th largest eigenvalue of a symmetric positive-definite matrix $M$ by $\lambda_j(M)$. Then can we say anything about convergence of $\lambda_j(A_n) \rightarrow \lambda_j(\Sigma)$ as $n \rightarrow \infty$, that is, whether it converges in probability or in distribution and if so can we characterize the rate of convergence.

Thanks a lot, any help is much appreciated.

Best Ashin

  • $\begingroup$ Do you mean $A_n = \frac{1}{n} \sum_{i=1}^n X_i X_i^T$? Also, presumably the $X_i$ are independent. $\endgroup$ – Robert Israel Dec 14 '11 at 18:51
  • 2
    $\begingroup$ There's a lot of literature on this problem. I suggest starting by taking a look at this paper: www-personal.umich.edu/~romanv/papers/sample-covariance.pdf and other recent papers by Roman Vershynin. $\endgroup$ – Mark Meckes Dec 14 '11 at 19:28
  • $\begingroup$ @ Robert : Yes I do mean $A_n = \frac{1}{n}\sum_{i = 1}^{n}X_i X_i^T$. Sorry about the typo. $\endgroup$ – Ashin Dec 14 '11 at 19:59
  • $\begingroup$ @Ashin: then perhaps you should edit the question accordingly. $\endgroup$ – Ori Gurel-Gurevich Dec 15 '11 at 0:45

I think this book is suitable for further study about this question:

"Representation Theory of Finite Groups An Introductory Approach" "Benjamin Steinberg"

This book has some chapters related to your question.


Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.