Proving Both Upper and Lower Bounds for Stirling's Approximation

  • Context: Comp Sci 
  • Thread starter Thread starter rxh140630
  • Start date Start date
Click For Summary
SUMMARY

The discussion centers on proving both upper and lower bounds for Stirling's approximation of factorials, specifically the expression n! = √(2πn)(n/e)^n(1 + θ(1/n)). The participants explore the necessity of proving bounds such as c₁nlgn ≤ lg(√(2πn)) + lg((n/e)ⁿ) + lg(1 + θ(1/n)) ≤ c₂nlgn, given different assumptions about lg(√(2πn)). The consensus is that both upper and lower bounds must be managed to ensure the accuracy of Stirling's approximation, supported by the formal definition of theta notation.

PREREQUISITES
  • Understanding of Stirling's approximation and its applications in combinatorics.
  • Familiarity with logarithmic functions and properties, particularly lg(n).
  • Knowledge of asymptotic notation, specifically theta (θ) notation.
  • Basic calculus concepts related to limits and bounding functions.
NEXT STEPS
  • Study the derivation and applications of Stirling's approximation in statistical mechanics.
  • Learn about the formal definitions and properties of asymptotic notations, including big O and little o.
  • Explore advanced topics in analysis, such as the use of limits in proving bounds.
  • Investigate the implications of Stirling's approximation in algorithm complexity analysis.
USEFUL FOR

Mathematicians, computer scientists, and students in advanced mathematics or algorithm design who are interested in factorial approximations and their applications in theoretical computer science.

rxh140630
Messages
60
Reaction score
11
Homework Statement
prove lg(n!) = thetha(nlgn) where lg n is log with base = 2. Use Sterling's approximation as a hint
Relevant Equations
Sterling's approximation: [itex]n! = \sqrt(2*pi*n)(\frac{n}{e})^n(1+ \theta{1/n}) [/itex]
Sterling's approximation: [itex]n! = \sqrt{2*pi*n}(\frac{n}{e})^n(1+ \theta(1/n))[/itex]

So I need to prove

[itex]c_1nlgn ≤lg(\sqrt{2*pi*n}) + lg((\frac{n}{e})^n) + lg(1+ \theta(1/n))) ≤ c_2nlgn[/itex]

My question is:

assume I've proven [itex]lg(\sqrt{2*pi*n})[/itex] as [itex]\theta(lgn)[/itex]

Do I need to now prove that [itex]c_1nlgn ≤ lgn ≤ c_2nlgn[/itex] ??

What if we assume I found that [itex]lg(\sqrt{2*pi*n})[/itex] as [itex]\theta(\sqrt{n})[/itex]

Do I need to now prove that [itex]c_1nlgn ≤ \sqrt{n} ≤ c_2nlgn[/itex] ??

Assuming the other terms in the sum [itex]g(\sqrt{2*pi*n}) + lg((\frac{n}{e})^n) + lg(1+ \theta(1/n)))[/itex] are proven to be [itex]\theta(nlgn)[/itex]
 
Physics news on Phys.org
The upper and lower bounds of Stirling's formula are so good, that I would use both of them just to be on the safe side.
$$
1< e^{1/(12n+1)} < \dfrac{n!}{\sqrt{2\pi n}\cdot\left(\dfrac{n}{e}\right)^n} < e^{(1/12n)}< 1+\dfrac{1}{11n}
$$

The formal definition of the theta notation is:
$$
f(x)=\theta(g(x)) \Longleftrightarrow 0<\operatorname{lim\,inf}_{x\to a}\left|\dfrac{f(x)}{g(x)}\right|\leq \operatorname{lim\,sup}_{x\to a}\left|\dfrac{f(x)}{g(x)}\right|<\infty
$$
so, yes, you have to manage both ends.
 

Similar threads

  • · Replies 11 ·
Replies
11
Views
2K
  • · Replies 3 ·
Replies
3
Views
3K
  • · Replies 15 ·
Replies
15
Views
1K
  • · Replies 13 ·
Replies
13
Views
2K
  • · Replies 6 ·
Replies
6
Views
550
Replies
24
Views
3K
  • · Replies 14 ·
Replies
14
Views
3K
  • · Replies 7 ·
Replies
7
Views
2K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 2 ·
Replies
2
Views
6K