3
$\begingroup$

After reading a bit on the wiki, I though I could find the variance of my variable, but I'm getting confused... Let's assume we have two random variables, $A$ and $B$ which are uncorrelated. Now, what is the variance of:

$D = f*(A-B)+B$

My problem is, from this formula I get:

$Var(D) = f^{2}(Var(A)+Var(B)) + Var(B)$

However, the initial definition of D can be rewritten as:

$D = f*A + (1-f)*B$

which in turns gives

$Var(D) = f^{2}Var(A) + (1-f)^{2}Var(B)$

Maybe I'm seeing wrong, but the two variance formulas are different

$\endgroup$
3
  • $\begingroup$ Is $f$ a constant? Does $*$ denote multiplication? $\endgroup$ Commented Aug 12, 2016 at 14:19
  • $\begingroup$ Yes, $f$ is constant and $*$ stands for multiplication $\endgroup$ Commented Aug 12, 2016 at 14:26
  • $\begingroup$ I think it is a bit confusing. $\cdot$ is better suited than * in this case. $f$ suggests a function, I would have used $c$ instead. But now that there already is an answer using the same notation, let is stay. $\endgroup$ Commented Aug 12, 2016 at 14:30

1 Answer 1

4
$\begingroup$

If $A$ and $B$ are uncorrelated, $A-B$ and $B$ certainly are correlated. Therefore, you cannot write:

$Var(D) = f^{2}(Var(A)+Var(B)) + Var(B)$

Instead:

$Var(D) = f^{2}(Var(A)+Var(B)) + Var(B) + 2fCovar(A-B,B)$ $Var(D) = f^{2}(Var(A)+Var(B)) + Var(B) - 2fVar(B)$

Which gives, after factorization :

$Var(D) = f^{2}Var(A) + (1-f)^{2}Var(B)$

$\endgroup$
3
  • $\begingroup$ Can you explain how you remove $Covar(A-B, B)$? $\endgroup$ Commented Aug 16, 2016 at 10:46
  • $\begingroup$ @DiogoSantos $Covar(A-B,B) = -Var(B)$ since $A$ and $B$ are uncorrelated $\endgroup$
    – RUser4512
    Commented Aug 16, 2016 at 10:48
  • $\begingroup$ But why does the variance of $A$ disappears? $\endgroup$ Commented Aug 16, 2016 at 13:14

Your Answer

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

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