Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I have two random variables $X$ and $Y$ which follows Normal distribution , whose pdf's are given by

$f(x)= \frac{1}{2 \sqrt{2 \pi} \sigma}[e^{\frac{-(x-1)^2}{2 \sigma^2}}+e^{\frac{-(x+1)^2}{2 \sigma^2}}]$

$f(y)= \frac{1}{ \sqrt{2 \pi} \sqrt{1+ \sigma^2}}[e^{\frac{-x^2}{2(1+\sigma^2)}}]$

Both Random Variables $X$ and $Y$ has Mean Zero and Variance $1+\sigma^2$.

I am interested to evaluate the Higher moments for both $X$ and $Y$.

Is the higher moments for both $X$ and $Y$ are equal. What is the difference between these two Random variables.

Thank you.

share|improve this question
Your first variable is not normal; it is a mixture of normals. – whuber Mar 11 '13 at 18:47
That's homework? – ziggystar Mar 11 '13 at 20:37
@ziggystar the homework tag now maps to self-study, with similar intent – Glen_b Mar 11 '13 at 23:24
up vote 3 down vote accepted

By reading off the arguments of the exponentials, it is evident that $X$ is a mixture of Normals centered at $\pm 1$, whereas $Y$ is a Normal with the same mean and variance as $X$. Here are their density functions when $\sigma=1/2$:


$X$ is blue with two peaks; $Y$ is red with one peak. (When $\sigma$ is much larger than $1/2$, $X$ will have only a single "merged" peak, but it will be flatter at the top than $Y$'s peak.)

From this picture it is apparent that

  • All odd moments will be zero, because both variables are symmetric about the origin.

  • The red curve ($Y$) has fatter tails than the blue ($X$), implying its higher even moments will be greater.

The easiest way to do the calculations is with the characteristic or moment generating functions. The MGF of a Normal$(\mu, \tau)$ distribution is

$$\exp{(t \mu +\frac{t^2 \tau ^2}{2})};$$

when expanded as a MacLaurin series in $t$, the coefficient of $t^n$ is $1/n!$ times the $n$th moment. Plugging in $\mu=0$ and $\tau = \sqrt{1+\sigma^2}$ gives the MGF of $Y$ as

$$1+\left(\frac{1}{2}+\frac{\sigma ^2}{2}\right) t^2+\left(\frac{1}{8}+\frac{\sigma ^2}{4}+\frac{\sigma ^4}{8}\right) t^4+\left(\frac{1}{48}+\frac{\sigma ^2}{16}+\frac{\sigma ^4}{16}+\frac{\sigma ^6}{48}\right) t^6+\left(\frac{1}{384}+\frac{\sigma ^2}{96}+\frac{\sigma ^4}{64}+\frac{\sigma ^6}{96}+\frac{\sigma ^8}{384}\right) t^8+\ldots$$

whereas that of $X$ is the average of MGFs of its components, which works out to

$$1+\left(\frac{1}{2}+\frac{\sigma ^2}{2}\right) t^2+\left(\frac{1}{24}+\frac{\sigma ^2}{4}+\frac{\sigma ^4}{8}\right) t^4+\left(\frac{1}{720}+\frac{\sigma ^2}{48}+\frac{\sigma ^4}{16}+\frac{\sigma ^6}{48}\right) t^6+\left(\frac{1}{40320}+\frac{\sigma ^2}{1440}+\frac{\sigma ^4}{192}+\frac{\sigma ^6}{96}+\frac{\sigma ^8}{384}\right) t^8+\ldots.$$

The agreement with the preceding MGF in the first two terms is evident. The coefficients of $t^4$ give the fourth moments: comparing, we see that of $Y$ exceeds that of $X$ by $4!(\frac{1}{8}-\frac{1}{24})\gt 0$, as expected. Similar comparisons bear out (and quantify) our impression that $Y$ has the greater higher moments.

share|improve this answer
That hand-drawn look on the axes and tick labels seems 'friendly'. Is that one the xkcd-style? – Glen_b Mar 11 '13 at 23:28
@Glen_b Yes. I am learning to distinguish graphics that need to be clear and precise from those that benefit from some "friendly" simplification. This, evidently, is one of the latter. – whuber Mar 11 '13 at 23:34
Thank you so much for explanation – Gold Mar 12 '13 at 9:25

Your Answer


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.