3
$\begingroup$

I'm trying to calculate first passage time distribution for some processes in Mathematica. First I tried to calculate this for Wiener process with zero drift and $\sigma = 1$ and the absorbing boundary at $a = 2%$. It's the well-known formula (Levy-Smirnov distribution) $${f_{FP}} = \frac{a}{{\sqrt {2\pi {\sigma ^2}{t^3}} }}{e^{ - \frac{{{a^2}}}{{2{\sigma ^2}t}}}}$$

I used this code

sample = Map[FirstCase[#, _?(Last[#] >= 2 &)] &, 
   RandomFunction[WienerProcess[0, 1], {0, 15, 0.001}, 10^4][
    "Paths"]];
data0 = DeleteCases[sample, {}][[All, 1]];
data = Cases[data0, _?Positive];
Show[Histogram[data, 40, "PDF"], 
 Plot[2/Sqrt[2*\[Pi]*t^3]*Exp[-(2^2/(2*t))], {t, 0, 15}, 
  PlotStyle -> Thick]]

but the result was strange enter image description here

I haven't got any idea why the simplest case failed. Could someone help me with this? Thanks in advance

P.S. I'm sorry for terrible English

$\endgroup$
  • 1
    $\begingroup$ FYI: InverseGammaDistribution[] is built-in, so you can do PDF[InverseGammaDistribution[1/2, a^2/(2 σ^2)], x]. $\endgroup$ – J. M. is away Oct 1 '17 at 18:19
  • $\begingroup$ Yes, but I want to understand, why the numeric results don't agree with exact formula, because I'm going to calculate FPT for non-Markovian process (and there is no exact formula). $\endgroup$ – Msdos4 Oct 1 '17 at 18:24
4
$\begingroup$

Perhaps you were expecting this:

pdf = 2/Sqrt[2*π*t^3]*Exp[-(2^2/(2*t))];
normalization = NIntegrate[pdf, {t, 0, 15}];
Show[Histogram[data, 40, "PDF"], 
 Plot[pdf/normalization, {t, 0, 15}, PlotStyle -> Thick]]

Mathematica graphics

$\endgroup$
  • $\begingroup$ The function being plotted here is effectively PDF[TruncatedDistribution[{0, 15}, InverseGammaDistribution[1/2, 2]], t]; that is, the expected heavy tail distribution with a good amount of the tail cut off. $\endgroup$ – J. M. is away Oct 1 '17 at 19:58
  • $\begingroup$ Thank you, I didn't think about normalization. But are there ways to normalize data (histogram), not function? $\endgroup$ – Msdos4 Oct 1 '17 at 20:31
  • 2
    $\begingroup$ @Msdos4 Yes, the third argument of Histogram, where you have "PDF", can take an arbitrary function. See the documentation "Details" section of Histogram. $\endgroup$ – Michael E2 Oct 1 '17 at 21:13
  • $\begingroup$ Thank you once again! Sorry for stupid questions, I'm not good in Mathematica but I want to learn. $\endgroup$ – Msdos4 Oct 1 '17 at 21:18
  • $\begingroup$ @Msdos4 You're welcome. :) $\endgroup$ – Michael E2 Oct 1 '17 at 21:21

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.