How to solve these equations? (from 10.12.c, D. Williams, Probability with Martingales)

54 Views Asked by At

In Sec. 10.12 (hitting times of simple random works), eq. (c) (page 103) of Probability with Martingales (D. Williams 1991), the author said:

For (any) $\theta\in \mathbb{R}$, let $\alpha:=\mathrm{sech}(\theta)$, we have: $$\sum_{n=1}^{\infty} \alpha^n \mathbb{P}(T=n)=\alpha^{-1}[1-\sqrt{1-\alpha^2}]$$ so that $$\mathbb{P}(T=2m-1)=(-1)^{m+1}\binom{\frac{1}{2}}{m}$$ $T$ is the stopping time of the simple random work, i.e., a discrete random variable taking $\{1,3,5,...\}$

How did he get such a result?

My conjecture: one may let (say) $\alpha=0,0.1,0.01,...,1 \in[0,1]$ to create infinitely many equations and solve each (of the infinite many) variable $\mathbb{P}(T=2m-1)$ for $m=1,2,3,...$ in these equations.

Besides, I do not really understand the notation $\binom{\frac{1}{2}}{n}$. If it denotes the combination number, how can one write a fraction $\frac{1}{2}$?

1

There are 1 best solutions below

0
On BEST ANSWER

For a real number $b$ and a non-negative integer $n$, the meaning of $b\choose n$ is $$ {b(b-1)\cdots(b-n+1)\over n!},\qquad n\ge 1, $$ with ${b\choose 0}=1$ being agreed upon by convention. With this notation, you have Newton's formula $$ (1+t)^b=\sum_{n=0}^\infty {b\choose n} t^n,\qquad |t|<1. $$ You can use this with $b=1/2$ and $t=-\alpha^2$ to expand $\alpha^{-1}\left[1-\sqrt{1-\alpha^2}\right]$ in a power series. Then equate coefficients of like powers of $\alpha$ to see that $$ P(T=2m-1) = (-1)^{m+1}{1/2\choose m},\qquad m=1,2,\ldots. $$