0
$\begingroup$

As the title already says I have to show that for $Z=(Z_1,...,Z_n)^T$ it is $$ y^T\text{Cov}(Z)y\geqslant 0~\forall~y\in\mathbb{R}^n. $$

In a book I read $$ y^T\text{Cov}(Z)y=\text{Var}(y^TZ)\geqslant 0 $$ but that's not clear to me, because I would have said that $$ y^T\text{Cov}(Z)y=\text{Cov}(y^TZ) $$ (and that can be negative). So in other words, I do not see why $$ \text{Cov}(y^TZ)=\text{Var}(y^TZ). $$


Symmetric is clear because of $\text{Cov}(Z_i,Z_j)=\text{Cov}(Z_j,Z_i)$.

$\endgroup$
1
$\begingroup$

Indeed $y^T\text{Cov}(Z)y=\text{Var}(y^TZ)$ (variance, not covariance) since $y^TZ$ is a random matrix of size $1\times1$, that is, a real-valued random variable. Thus, $y^T\text{Cov}(Z)y$ is a nonnegative real number for every vector $y$, which proves that the symmetric matrix $\text{Cov}(Z)$ of size $n\times n$ is nonnegative definite.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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