Consider a random variable X with the following moment generating function:


(a) Find the expected value of $3 - X$.

(b) Find variance of $3 - X$.

What I have reached till now is that is (most probably?) geometric distribution which means that the probability is 0.2, and that if we want to get the expected value of $3 - X$ then it would be equal to $E(3-X) = E(3) - E(X)$

What I think: E(X) = 1/0.2 But I have no idea where to go from there.


a) $E(3-X) = E(3) - E(2Y) = E(3) - 2 * E(Y) = 3 - 2 * 1/0.2 = -7$ (I also used differentiation rule to double check, but differentiation is way longer)


I tried to use a similar logic to solve (b), since the variance = $p/(1-p)^2$ I thought I could equate $Var(X) = Var(2Y) = 2 * (0.8/0.2^2) = 40$

Not sure if this method is right or not, though..


Answer of B: Okay, so, $Var(X) = Var(2Y) = 2^2 Var(Y) = 4 * (0.8/0.2^2) = 80$

Thank you all! :)

  • $\begingroup$ Hint: The expected value of a constant is that constant. $\endgroup$ – Mosquite Oct 18 '16 at 18:26
  • $\begingroup$ Your edit is right. $\endgroup$ – callculus Oct 18 '16 at 18:55
  • $\begingroup$ You can proof your result of the variance by using differentiations. $\endgroup$ – callculus Oct 18 '16 at 19:07
  • $\begingroup$ The variance of a translation of a random variable is the same as that for the random variable. So $Var(3-X)=Var(X)$. $\endgroup$ – Paul Oct 18 '16 at 19:10
  • $\begingroup$ Yeah. $Var(X)=80$ is right. $\endgroup$ – callculus Oct 18 '16 at 19:16

The property of moment-generating functions that you should recall for this problem is $$\left[\frac{\partial^k M_X(t)}{\partial t^k}\right]_{t=0} = M^{(k)}_X(0) = \operatorname{E}[X^k];$$ that is to say, the $k^{\rm th}$ derivative of the MGF at $t = 0$ is the $k^{\rm th}$ raw moment of $X$, whenever such a moment is defined. So, for $k = 1$, we observe that the first derivative of $M_X$ at $t = 0$ gives the expectation; for $k = 2$, we get the expectation of $X^2$. Then to obtain the variance, you would calculate $$\operatorname{Var}[X] = \operatorname{E}[X^2] - \operatorname{E}[X]^2.$$

To understand where this relationship comes from, recall that the MGF is defined as $$M_X(t) = \operatorname{E}[e^{tX}] = \operatorname{E}\left[\sum_{k=0}^\infty \frac{(tX)^k}{k!}\right] = \sum_{k=0}^\infty \frac{\operatorname{E}[X^k]}{k!} t^k.$$ But by Taylor's theorem, $$M_X(t) = \sum_{k=0}^\infty \frac{M_X^{(k)}(0)}{k!} t^k,$$ thus $\operatorname{E}[X^k] = M_X^{(k)}(0)$ as claimed (again, whenever the moments are defined).


If $Y$ is geometric with success probability $0.2$, then $m_Y(t)=\frac{0.2 e^t}{1-0.8 e^t}$. You can show that then $X=2Y$ has your given MGF $m_X(t)=m_Y(2t)=m_{2Y}(t)$.


Hint: Differentiate $m(t)=\large{\frac{0.2\cdot e^{2\cdot t}}{1-0.8\cdot e^{2\cdot t}}}$ w.r.t. $t$. For this purpose apply the quotient rule.

Then set $t=0$ to get the expected value. In short form


And the variance is $Var(X)=m''(0)-[m'(0)]^2$ and $Var(3-X)=Var(X)$

  • $\begingroup$ No. We have $\operatorname{Var}[X] = m''(0)$ if and only if $\operatorname{E}[X] = 0$. $\endgroup$ – heropup Oct 18 '16 at 18:42
  • $\begingroup$ Thanks for the comment, I have made an edit. $\endgroup$ – callculus Oct 18 '16 at 18:44

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.