3
$\begingroup$

The question I'm trying to solve is the following.

Independent trials, each resulting in success with probability p, are performed until $k$ consecutive successful trials have occurred. Let $X$ be the total number of successes in these trials, and let $P_n = P(X = n)$. Find $P_k$.

The way I understand the question is that I need to find the probability that, if we perform $m$ trials, there will be $k$ successes, and they will be consecutive.

So I did the following:

$$ P_k = \sum_{i=0}^{\infty}P(X=k|F=i)P(F=i) $$ F is the number of failures at the beginning. $$ P_k = \sum_{i=0}^{\infty}P(X=k)P(F=i) $$ $$ P_k = \sum_{i=0}^{\infty}p^k(1-p)^i $$ $$ P_k = p^k\sum_{i=0}^{\infty}(1-p)^i $$ $$ P_k = \frac{p^k}{1-p} $$

However, the answer is $p^{k-1}$ and I'm struggling to understand why.

$\endgroup$
2
  • $\begingroup$ Welcome to stackexchange. You seem to describe three different problems here. The title says you want to find the probability of getting only k successes, the next description describes a geometric random variable, and your interpretation introduces a variable $m$ that was not defined or described. You'll need to clarify the problem if you hope to get a good answer. Best of luck! $\endgroup$
    – jtb
    Feb 3 at 21:09
  • $\begingroup$ What is the source of this problem? A very similar problem was asked a few minutes ago. $\endgroup$
    – lulu
    Feb 3 at 21:53

2 Answers 2

3
$\begingroup$

The only way $X$ could be $k$ is if the string of outcomes had the form $F^mS^k$ where $F$ denotes Fail, $S$ denotes Success, and $m$ could be any non-negative number.

Now $P(F^mS^k)=(1-p)^mp^k$ so $$P(X=k)=p^k\sum_{m=0}^{\infty} (1-p)^m=p^k\times \frac 1{1-(1-p)}=p^{k-1}$$as desired.

$\endgroup$
0
2
$\begingroup$

To get $X=k$:

  • you need to have the first success preceded by any non-negative number of failures (with probability $1$ that you reach this state at some stage)

  • followed immediately by $k-1$ more successes, so with probability $p^{k-1}$.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .