1
$\begingroup$

I'm aware that $$Var[A+B] = Var[A] + Var[B] + 2Cov[A,B]$$ and $$Var[2C] = 4Var[C].$$ So then $$Var[A+B-2C] = Var[A] + Var[B] + 4Var[C] + 2Cov[A,B] - 4Cov[B,C] - 4Cov[A,C].$$ Is this correct?

$\endgroup$
1
$\begingroup$

It is called the Bilinearity of Covariance.

$$\begin{align}\mathsf{Cov}\left(\raise{0.75ex}{\sum_{i\in\mathcal I} a_iX_i,\sum_{j\in\mathcal J} b_jY_j}\right)&=\sum_{\langle i,j\rangle\in\mathcal {I{\times}J}} a_ib_j~\mathsf{Cov}(X_i,Y_j)\end{align}$$

In the specific case of variance

$$\begin{align}\mathsf{Var}\left(\raise{0.75ex}{\sum_{i\in\mathcal I} a_iX_i}\right)&=\sum_{\langle i,j\rangle\in\mathcal {I^2}} a_ia_j~\mathsf{Cov}(X_i,X_j)\\[1ex]&=\sum_{i\in\mathcal I} a_i^2~\mathsf{Var}(X_i)+2~\mathop{\sum\qquad}_{\langle i,j\rangle\in\mathcal I^2:i<j}~a_i a_j~\mathsf{Cov}(X_i,X_j)\end{align}$$

So indeed$${\mathsf{Var}(A+B-2C)}={{\mathsf{Var}(A)+\mathsf{Var}(B)+4~\mathsf{Var}(C)}+{2~\mathsf{Cov}(A,B)-4~\mathsf{Cov}(A,C)-4~\mathsf{Cov}(B,C)}}$$

$\endgroup$

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.