X is a random variable with a pdf of $ f(x) = \begin{cases} x/2, & 0 \le x \le 2 \\ 0, & \text{otherwise} \end{cases}$

I tried finding the characteristic function of this but ended up with an expression where $\phi(0)\neq 1 $ and I cant pinpoint why as I've followed the process exactly (Edited to implement gunes comment - still seem to have the same problem though):

$\phi(t) = E[e^{itX}]$

$\qquad= \int_0^2 e^{itx}\cdot\frac{1}{2}x \quad dx $

Integrate by parts: $u = x, \quad u'=1,\quad v'=e^{itx},\quad v=\frac{e^{itx}}{it}$

$\phi(t)= \frac{1}{2} \Big( \Big[\frac{x}{it}e^{itx}\Big]_0^2 \quad -\quad \int_0^2 \frac{1}{it}e^{itx}dx \Big) $

$\qquad= \frac{1}{2} \Big( \frac{2}{it}e^{2it} \quad -\quad \Big[\frac{1}{(it)^2}e^{itx}\Big]_0^2 \quad \Big) $

$\qquad= \frac{1}{2} \Big( \frac{2}{it}e^{2it} \quad -\quad \Big(\frac{1}{(it)^2}e^{2it} - \frac{1}{(it)^2}\Big) \Big) $

$\qquad= \frac{1}{2} \Big( \frac{it\cdot2e^{2it}-e^{2it}+1}{(it)^2} \Big) $

Edit2: Applying L'Hopital's Rule to verify $\, \phi(0) = 1$:

$\lim\limits_{t \to 0} \frac{e^{2it}-2ite^{2it}-1}{2t^2} \quad=\quad \lim\limits_{t \to 0} \frac{2ie^{2it}-2ie^{2it}+4te^{2it}}{4t} \quad=\quad \lim\limits_{t \to 0}e^{2it} = 1$

Thanks for the help =)

  • $\begingroup$ Besides the error mentioned in the comments, your derivation assumes $t \neq 0$. That is why you get contradicting results. If you assume $t = 0$ it is pretty easy and $\phi(0) = \mathbb E[\exp(0)] = \mathbb E[1] = 1$. $\endgroup$ – Syd Amerikaner May 10 '19 at 10:54
  • $\begingroup$ Yep, I mainly use the $\phi(0)=1$ result as a check that my expansion is correct. As you pointed out it holds if you substitute it straight in but expanding the characteristic function will allow me to find $E[X]$ and $Var[X]$ next which is why I want to see why it no longer holds in the end $\endgroup$ – Maharero May 10 '19 at 13:04
  • 2
    $\begingroup$ I think you got my point wrong. You did everything (besides the error correct). This, in fact, is a characteristic function and it is the characteristic function for your density. More clearly, the characteristic function for this density is defined as $$\phi(t) = \begin{cases} \frac{\exp(2it) - 2it\exp(2it) - 1}{2t^2} & \text{if $t\neq 0$ and} \\ 1 & \text{if $t = 0$} \end{cases}$$ $\endgroup$ – Syd Amerikaner May 10 '19 at 13:56
  • 1
    $\begingroup$ Although your expression for $\phi$ may be undefined, its value is not. You may obtain it either directly, by re-evaluating the integral with $t$ explicitly set to $0,$ or by applying L'Hopital's Rule. $\endgroup$ – whuber May 10 '19 at 14:38

You should have calculated the integral: $$\phi(t)=E[e^{itX}]=\int_0^2 e^{itx}\frac{x}{2} dx$$

But, instead in the second line you assert that $E[e^{itX}]=E[e^{itx/2}]$, which is wrong, because in general: $$E[e^{itX}]\neq E[e^{itf(x)}]$$

Edit: Based on your correction, your final characteristic function seems OK. We just need to evaluate the limit when $t\rightarrow 0$, which can be done via L'Hopital Rule.

| cite | improve this answer | |
  • $\begingroup$ Makes sense - I've updated my question with your answer however I seem to still reach the same problem in the end $\endgroup$ – Maharero May 10 '19 at 13:08
  • $\begingroup$ @Maharero Looks like just a simple mix-up... second integral should be $v \hspace{1mm} u'$ $\endgroup$ – compbiostats May 10 '19 at 13:26

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.