1
$\begingroup$

I have a function of the form:

$f(n) = \prod_{i=1}^{n-1} (1-ai)$

Here, $a \geq 0$ and $(a*i) < 1$. For $n > 10^5$ or $10^6$, what is the best possible analytic approximation for $f(n)$ that will allow me compute values for the function with reasonable computational resources? What bounds on the error of the approximation can I expect?

$\endgroup$
4
  • $\begingroup$ What is the role of $i$ in the product? $\endgroup$ Aug 15, 2012 at 17:09
  • $\begingroup$ @András Bátkai, that was incredibly sloppy, sorry about that. I've fixed f(x). $\endgroup$
    – Roger S.
    Aug 15, 2012 at 17:11
  • $\begingroup$ You could group terms in pairs to get terms like (1 - (n+1)a + ba^2). For small values of a this might help you with your error estimates. Gerhard "Ask Me About System Design" Paseman, 2012.08.15 $\endgroup$ Aug 15, 2012 at 17:15
  • $\begingroup$ @Gerhard Paseman, thanks for the good suggestion! I've been trying things a bit like that, but the trouble is that $a$ isn't necessarily small, and I'm looking for some error bounds for the approximation. $\endgroup$
    – Roger S.
    Aug 15, 2012 at 17:40

1 Answer 1

5
$\begingroup$

$$f(n) = \dfrac{(-a)^{n-1} \Gamma(n - 1/a)}{\Gamma(1 - 1/a)}$$ If $a n < 1$, you'll want to use the reflection principle $$\Gamma(1-z) \; \Gamma(z) = {\pi \over \sin{(\pi z)}}$$ so $$ f(n) = \dfrac{a^{n-1} \Gamma(1/a)}{\Gamma(1-n+1/a)}$$

Use Stirling's series for the asymptotic approximation of $\ln(\Gamma(1-n+ 1/a))$, or one of its variants that you can find at http://en.wikipedia.org/wiki/Stirling%27s_approximation

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.