integration by parts with normal and complex

75 Views Asked by At

Hey could someone help me to compute the following integrals, I have two results in front of me, so maybe it would be helpful to see your steps! thanks a lot: We know that the Euler scheme of a forward SDE is given as follows, \begin{align*} X_{i+1}^\pi= X_{i}^\pi + b \Delta t + \sigma \Delta W_i \end{align*}

Suppose we want to approximate the expectation we need to approximate the expectations of the form: \begin{equation} \label{FCB-1} I_0:= \mathbb{E}\left [v(t_{i+1},X_{i+1}^{\pi})\Delta W_i|X_{i}^{\pi}=x \right] \end{equation}
where $v$ represents a general function.\ In the next step we use the Fourier-Cosine method on $v(t_{i+1},\xi)$, we get \begin{align*} v(t_{i+1},\xi)&= \sum_{k=0}^{\infty} \text{'} V_k \cos \left(k \pi \frac{\xi -a}{b-a} \right)\\ &\approx \sum_{k=0}^{N} \text{'} V_k \cos \left(k \pi \frac{\xi -a}{b-a} \right) \end{align*} This gives us the following approximation, \begin{align*} & I_1= \sum_{k=0}^{N} \text{'} V_k \mathbb{E} \left[ \cos \left(k \pi \frac{\xi -a}{b-a}\right) \Delta W_i\bigg \lvert X_i^{\pi }=x \right] \\ \text{with }& V_k(t_{i+1})=\frac{x}{b-a} \int_a^b v(t_{i+1},\xi)\cos \left(k \pi \frac{\xi -a}{b-a} \right) d \xi. \end{align*} From $I_1$ we will rewrite the expectation, \begin{align*} \mathbb{E} &\left[ \cos \left(k \Pi \frac{\xi -a}{b-a}\right) \Delta W_i \bigg \lvert X_i^{\pi }=x \right]\\ &= \mathcal{R}\left \{ \mathbb{E}\left[exp \left( ik \pi \frac{x+ b\Delta t+ \sigma \Delta W_i-a}{b-a} \right) \Delta W_i \bigg \lvert X_i^{\pi }=x \right] \right\}\\ &= \mathcal{R}\left \{ \mathbb{E}\left[exp \left( ik \pi \frac{x+ b\Delta t-a}{b-a} \right) \Delta W_i exp \left( ik \pi \frac{ \sigma \Delta W_i}{b-a} \right) \right] \right\}\\ &= \mathcal{R}\left \{ exp \left( ik \pi \frac{x+ b\Delta t-a}{b-a} \right)\mathbb{E} \left[ \Delta W_i exp \left( ik \pi \frac{ \sigma \Delta W_i}{b-a} \right) \right] \right\} \end{align*} We focus on the last term \begin{align*} \mathbb{E} \left[ \Delta W_i exp \left( ik \pi \frac{ \sigma \Delta W_i}{b-a} \right) \right] = \frac1{\sqrt{2 \pi \Delta_t}} \int_{- \infty}^{\infty} y\, \text{exp}\left(ik\pi \frac{\sigma y}{b-a}\right)\, \text{exp}\left(- \frac{y^2}{\Delta_ t}\right)\,\mathrm d y \end{align*} and from there i am stuck :S, the $\Delta W_i \sim N(0,\Delta t) $ The final result/approximation of I_0 should be $\sum_{k=0}^{N-1}V_k(t_{i+1})\sigma \Delta t \mathcal{R}\left( i \frac{k \pi}{b-a}\phi(\frac{k \pi}{b-a})exp {ik\pi \frac{x-a}{b-a }}\right)$

1

There are 1 best solutions below

0
On

Starting with

$$I = \frac1{\sqrt{2 \pi \Delta_t}} \int_{- \infty}^{\infty} y\, \text{exp}\left(ik\pi \frac{\sigma y}{b-a}\right)\, \text{exp}\left(- \frac{y^2}{\Delta_ t}\right)\,\mathrm d y$$

make the following substitutions

$$y = \sqrt{\pi \Delta_t} x$$ $$s = k \dfrac{\sqrt{\pi \Delta_t}}{2}\dfrac{\sigma}{a-b}$$

and rewrite the integral such that it is easily recognizable as a Fourier Transform

$$I = \dfrac{\sqrt{\pi \Delta_t}}{\sqrt{2}}\int_{- \infty}^{\infty} x\, \text{exp}\left(-\pi x^2\right)\, \text{exp}\left(-2\pi i xs\right)\,\mathrm d x$$

$$ = -\dfrac{1}{2\pi i} \dfrac{\sqrt{\pi \Delta_t}}{\sqrt{2}}\int_{- \infty}^{\infty} -2\pi ix\, \text{exp}\left(-\pi x^2\right)\, \text{exp}\left(-2\pi i xs\right)\,\mathrm d x$$

$$ = -\dfrac{1}{2\pi i} \dfrac{\sqrt{\pi \Delta_t}}{\sqrt{2}}F\left\{-2\pi ix\, \text{exp}\left(-\pi x^2\right)\right\}$$

so by the derivative property of the Fourier Transform

$$ I = -\dfrac{1}{2\pi i} \dfrac{\sqrt{\pi \Delta_t}}{\sqrt{2}} \dfrac{\mathrm d}{\mathrm{d}s}F\left\{\text{exp}\left(-\pi x^2\right)\right\}$$

and looking up the Fourier Transform of the Gaussian in a table of Fourier Transforms

$$ I = -\dfrac{1}{2\pi i} \dfrac{\sqrt{\pi \Delta_t}}{\sqrt{2}} \dfrac{\mathrm d}{\mathrm{d}s}\left[\text{exp}\left(-\pi s^2\right)\right]$$

$$ = -i \dfrac{\sqrt{\pi \Delta_t}}{\sqrt{2}} s \, \text{exp}\left(-\pi s^2\right)$$

and then finally substituting back in for $s$ and performing a little algebra

$$ I = i \dfrac{\pi \Delta_t}{2\sqrt{2}}\dfrac{\sigma}{b-a} k \, \text{exp}\left(-\dfrac{\pi^2}{4}\left[\dfrac{\sigma}{b-a}\right]^2 \Delta_t k^2\right)$$

You should double check the algebra on that last step for mistakes.