In a research problem (related to statistics), I've encountered a k-dimensional Dirichlet integral of the form:(adsbygoogle = window.adsbygoogle || []).push({});

[tex]f_k(x) = \int_{x_c}^\infty dx_1 \ldots \int_{x_c}^\infty dx_k\, \delta\left(x-\left(x_1+\ldots+x_k\right)\right)\ x_1^{-s} \ldots x_k^{-s}[/tex]

where [tex]s>0[/tex]. Essentially I am integrating the function [tex]x_1^{-s} \ldots x_k^{-s}[/tex] over the plane defined by [tex]x_1+\ldots+x_k=x[/tex], with the criterion [tex]x_k \geq x_c[/tex] providing the necessary cutoff so that the integral converges.

While I have been able to write a program to numerically integrate this, I am wondering if there is a closed form solution. It seems that this integral is related in some way to a k-dimensional generalization of the incomplete beta function or an integral of a Dirichlet probability distribution. However, when trying to generalize some of the derivations I've found in the literature, I find I run into trouble because of the cutoff (the integral is usually done for [tex]s<0[/tex]). I've been able to derive

[tex]f_2(x) = x^{(-2(s-1)-1)}\, \theta\left(x-2x_c\right)\, \beta\left(\frac{x_c}{x},1-\frac{x_c}{x},1-s,1-s\right)[/tex] (where [tex]\theta(x)[/tex] is the Heaviside step function and [tex]\beta(x)[/tex] is the generalized incomplete beta function)

but have been unable to generalize the canonical derivations to higher k. It seems like the answer should look something like:

[tex]f_k(x)=x^{(-k(s-1)-1)}\, \theta\left(x-kx_c\right)\, g_k\left(\frac{x_c}{x},s\right)[/tex]

where [tex]g_k\left(\frac{x_c}{x},s\right)[/tex] is the desired generalization.

Does anybody here have experience with these type of statistical integrals, or could point me to some literature? It seems like this problem must be solved, but I don't have much experience with these sorts of problems. I am also wondering if the approach taken in http://www.jstor.org/sici?sici=0002-9890(200102)108:2<151:TMVO>2.0.CO;2-B", which evaluates the integral by a sum of the integrand evaluated at the vertices of the plane, can be applied to this problem (I suspect that it cannot?) Thanks.

**Physics Forums | Science Articles, Homework Help, Discussion**

Dismiss Notice

Join Physics Forums Today!

The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

# Help with multidimensional integral (Dirichlet type)

Can you offer guidance or do you also need help?

Draft saved
Draft deleted

Loading...

Similar Threads for Help multidimensional integral |
---|

B Need some help with understanding linear approximations |

I Help with simplifying series of hyperbolic integrals |

I Help with understanding inexact differential |

I Help please with biocalculus question involving differentiation |

I Please help with nth derivative question |

**Physics Forums | Science Articles, Homework Help, Discussion**