0
$\begingroup$

$\varphi _{X}(w)=\sum ^{\infty}_{k=0}(-1)^k(wb)^{2k}$,and the mth moment of X is $E[X^m]=(-j)^m\frac{d^m \varphi _{X}(w)}{dw^m}=(-j)^m\sum_k (-1)^k \times 2k \times(2k-1)\times ... \times(2k-(m-1))\times b^{2k}w^{2k-m}$

From the formula we derive,we found the $2k=m$,and the solution said so $E[X^m]=(-j)^m \times m! \times b^m$,i know where are $(-j)^m$ and $m!$ and $b^m$ from and the reason that $w^{2k-m}$ is dissappeared,but where is the $\sum_k(-1)^k $ going?

$\endgroup$
1
$\begingroup$

Your formula to compute the $m^{th}$ moment is not correct. I assume that $j$ is defined such that $j^2 = -1$ (which is not the standard notation by the way), that $b$ is some positive constant, and that $\phi_X(w) = E[e^{jwX}]$ is well defined for $|w|<\frac{1}{b}$.

To answer your question, the correct formula is: $$ E[X^m] = (-j)^m \frac{d^m}{dw^m} \bigg|_{w=0} \phi_X(w) $$ i.e. you have to evaluate the $m^{th}$ derivative at $w=0$ to get the $m^{th}$ moment. If you are careful with your summation terms, you will see that in general the $m^{th}$ derivative will look like $$ \sum_{k = p}^{+\infty} (-1)^k \frac{(2k)!}{(2k-2p)!}b^{2k}w^{2k-2p}, \quad m=2p $$ or $$ \sum_{k = p+1}^{+\infty} (-1)^k \frac{(2k)!}{(2k-2p-1)!}b^{2k}w^{2k-2p-1}, \quad m=2p+1 $$ So evaluating the formula at $w=0$ makes all the terms disappear in the case $m=2p+1$ (all the terms contains a $w$ to some positive power) $$ E[X^m] = 0, \quad m \; odd $$ and for $m=2p$, all the terms except the first one i.e. $m=2k$ contains terms with $w$ to some positive power. So at $w=0$, only the first term is left and you get $$ E[X^m] = (-1)^{m/2} (-j)^m m! \: b^{m}, \quad m \; even $$

$\endgroup$
  • $\begingroup$ a little confused,why is $E[X^m]$ =0 when m is odd? $\endgroup$ – Shine Sun May 15 '18 at 13:27
  • $\begingroup$ The terms in the series all have $w^n$ for $n \geq 1$, so at $w=0$ they all vanish. Try it for $m=1$ by taking 1 derivative from the characteristic function you gave me. $\endgroup$ – monty47 May 15 '18 at 14:00
  • $\begingroup$ but this only goes to that when m=1,E[X]=0,not m=odd,E[X]=0 $\endgroup$ – Shine Sun May 15 '18 at 22:23
  • $\begingroup$ That was an example, my general proof shows that $E[X^m]=0$ for any m odd (morally for the same reasons as the case m=1). Is there anything you do not understand in the proof that I can clarify? $\endgroup$ – monty47 May 15 '18 at 22:51
  • $\begingroup$ but if m=3,the $ E[X^3]=(-1)^{3/2}(-j)^33!b^3$,why is it 0? $\endgroup$ – Shine Sun May 15 '18 at 23:00

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.