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.

The integration function is very complicated, as below, where $\theta, k \in \mathbb{R}_{+}$ $$\int_{-\infty }^{\infty }\frac{{e}^{-i\,t\,x}\,\left(1- {e}^{200\,i\,t}\right) \,{\left( 1-i\,t\,\theta\right) }^{k}\,\left(1-{e}^{i\,t\,x}\right)}{{t}^{2}}dt$$ There ara so many multiplier in the integration, and result should be formed as a function to $x$, and $\theta,k$ as it's parameters. I have no idea about that now.

btw, this is not a home work. it is originate from my question about extreme value of my custom distrbution which is a sum of gamma distribution and uniform distribution, described here.

Above integration is from probablity theory. I multiply characteristic functions of gamma distribution and uniform distribution, and inverse the result function to get distribution function of new distribution. $\theta$ and $k$ have to be estimated from samples.

I checked another inverse method to convert characteristc function to distribution function the integration can also be expressed as below. $$\int_{0}^{\infty }\frac{\left( \left( \mathrm{cos}\left( 200\,t\right) -1\right) \,\mathrm{sin}\left( k\,\mathrm{atan}\left( t\,\theta\right) \right) -\mathrm{sin}\left( 200\,t\right) \,\mathrm{cos}\left( k\,\mathrm{atan}\left( t\,\theta\right) \right) \right) \,{\left( {t}^{2}\,{\theta}^{2}+1\right) }^{\frac{k}{2}}\,\mathrm{sin}\left( t\,x\right) }{{t}^{2}}dt$$+$$\int_{0}^{\infty }\frac{\left( \left( 1-\mathrm{cos}\left( 200\,t\right) \right) \,\mathrm{cos}\left( k\,\mathrm{atan}\left( t\,\theta\right) \right) -\mathrm{sin}\left( 200\,t\right) \,\mathrm{sin}\left( k\,\mathrm{atan}\left( t\,\theta\right) \right) \right) \,{\left( {t}^{2}\,{\theta}^{2}+1\right) }^{\frac{k}{2}}\,\mathrm{cos}\left( t\,x\right) }{{t}^{2}}dt$$

it is a little bit more complex than the one before.

share|improve this question
    
Can you confirm or deny if that middle term is actually supposed to be $(1-e^{it\theta})^k$? Otherwise for $k$ not even that large the integral will diverge (and have no principal value either). –  anon May 5 '12 at 15:03
    
sure, i can add derivation as –  Readon Shaw May 5 '12 at 15:11
add comment

1 Answer

Not an answer, but too big for a comment.
Since you hinted at the probabilistic origin of the integral in question, it is likely that $k$ is actually negative, since the characteristic function of the $\Gamma(k,\theta)$ random variable is $\phi_X(t) = \left(1- i t \theta\right)^{-k}$. Then, for $U_1$ and $U_2$ independent uniformly distributed random variables on intervals $(0,d_1)$ and $(0,d_2)$, and both independent on $X$, and for r.v. $Z = X + U_1 + U_2$ we get: $$ \phi_Z(t) = \phi_{X}(t) \phi_{U_1}(t) \phi_{U_2}(t) = \exp\left(i \frac{d_1+d_2}{2} t\right) \operatorname{sinc}\left(\frac{ d_1 t}{2} \right) \operatorname{sinc}\left(\frac{ d_2 t}{2} \right) \left(1 - i t \theta\right)^{-k} $$ Now, you would like to compute the probability density function $f_Z(z)$ by means of the inverse Fourier transform: $$ f_Z(z) = \frac{1}{2 \pi} \int_{-\infty}^\infty \mathrm{e}^{-i t z} \phi_Z(t) \mathrm{d} t $$ The you seem to be interested in $f_Z(d_2)$. The answer will depend on whether $d_1 < d_2$, or $d_1 > d_2$.

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.