Bernstein Polynomials and Expected Value

468 Views Asked by At

The first equation in this paper

http://www.emis.de/journals/BAMV/conten/vol10/jopalagzyl.pdf

is:

$$\displaystyle B_nf(x)=\sum_{i=0}^{n}\binom{n}{i}x^i(1-x)^{n-i}f\left(\frac{i}{n}\right)=\mathbb E f\left(\frac{S_{n,x}}{n}\right)$$

where $f$ is a Lipschitz continuous real function defined on $[0,1]$, and $S_{n,x}$ is a binomial random variable with parameters $n$ and $x$.

How is this equation proven?

1

There are 1 best solutions below

0
On BEST ANSWER

$$\mathbb E(g)=\sum_{i}\mathbb P\left(S_{n,k}\right)g(i)$$

where:

$$g(S_{n,x})=f\left(\frac{S_{n,x}}{n}\right)$$

Therefore we have:

$$\mathbb E\left(f\left(\frac{S_{n,x}}{n}\right)\right)=\sum_{i=0}^{n}\mathbb P(S_{n,x}=i)f\left(\frac{i}{n}\right)$$

But $$\mathbb P(S_{n,x}=i)=\binom{n}{i}x^i(1-x)^{n-i}$$

Therefore

$$\sum_{i=0}^{n}\binom{n}{i}x^i(1-x)^{n-i}f\left(\frac{i}{n}\right)=\mathbb E\left(f\left(\frac{S_{n,x}}{n}\right)\right)$$ which was the original statement.