Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $X$ be a continuous random variable with a density $$f_X(x) = \begin{cases}e^{-(x-1)}& \text{for }x > 1\\ 0 & \text{otherwise}\end{cases}$$ How do I compute the moment generating function of X? What is the range on which the moment generating function is defined?

How do I used the moment generating function to compute the mean, the second moment and the variance of X?

share|improve this question
add comment

3 Answers 3

up vote 3 down vote accepted
+50

As noted above (and standard anyway) $M(t)=E(e^{tx})=\int_1^{oo} e^{tx}e^{-(x-1)} dx$. Note that if t=1 the integrand becomes simply e^1, and the integral does not exist. So M(1) is undefined. Also note that if $t>1$ the integrand is $e^{(t-1)x+1}>e^{(t-1)x}$ which since $t>1$ goes to infinity as $x$ does. So M(t) does not exist for t>1.

We claim now that M(t) exists for t<1. The antiderivative of the integrand is $\frac {e^{-x(1-t)}e^1}{t-1}$, which goes to $0$ as $x$ approaches infinity, and evaluates to $\frac {e^t}{t-1}$ at $x=1$. We thus have the formula

$M(t)=\frac {e^t}{1-t}$

for the moment generating function, on the domain $\{t:t<1\}$. Note that this domain includes an open neighborhood of 0, so that moment generating calculations of moments can be done as usual.

To get the mean, second moment, and variance we only need a few terms of the series for $M(t)$. We can get these by multiplying the series for $e^t$ by the series for $1/(1-t)$. That is,

$(1+t+t^2/2+t^3/6+...)(1+t+t^2+t^3+...)$

which gives first few terms $M(t)=1+2t+(5/2)t^2+(8/3)t^3+...$. Then the usual procedure of taking derivatives and putting in zero gives $E(X)=2$ and $E(X^2)=5$. Finally, using the standard formula for variance we get $V(X)=E(X^2)-[E(X)]^2=5-2^2=1$.

share|improve this answer
1  
If my series for $M(t)$ is right, first derivative is $2+5t+...$ and plugging in 0 gives 2 for $M'(0)$. You can say that $M(0)=1$ from my series, but that just means $E(X^0)=1$, true for any random variable. –  coffeemath Oct 10 '12 at 19:33
1  
You can also calculate the moments directly, by integrating power of $x$ times pdf. I just did this on maple and got the same answers, $E(X)=2$ and $E(X^2)=5$. This could also be done by hand, using integration by parts. –  coffeemath Oct 10 '12 at 19:47
add comment

The moment generating function is $$M_X(t) = E[e^{tX}] = \int_1^\infty e^{tx} e^{-(x-1)}\ dx$$ Hint: in order for this improper integral to be defined, you want to check what happens as $x \to \infty$.

share|improve this answer
    
As $x \to \infty$, $M_x(t) \to 0$? –  idealistikz Oct 10 '12 at 6:13
    
$x$ is the dummy variable in the integration defining $M_X(t)$. The question is, for what values of $t$ does that integral converge? –  Robert Israel Oct 10 '12 at 6:36
    
The integral converges when $t=0$. Can we assume $t<1$? –  idealistikz Oct 10 '12 at 6:41
    
I don't understand because as $x \to \infty$, $M_x(t) \to \infty$ if $t>1$. What am I missing? –  idealistikz Oct 10 '12 at 6:44
    
There is no such thing as $M_x(t)$. If $t > 1$, $e^{tx} e^{-x-1} \to \infty$ as $x \to \infty$, and then the integral diverges. Also if $t=1$, the integral diverges. –  Robert Israel Oct 10 '12 at 18:24
add comment

MGF is defined as $M_X(t)=E(e^{tX})$.

Calculation of moments can be done through $E(X^n)=\frac{d^n M_X}{dt^n}(0)$.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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