Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $X_1, X_2,\dots,X_n$ be $n$ i.i.d. geometric distributed random variables with successful probability $p$, that is $$P(X_i=k)=(1-p)^{k-1}p$$

Let $Y_i=X_i-\mathbb{E}(X_i) = X_i-\frac{1}{p}$, so $Y_1,\dots,Y_n$ are also i.i.d. random variables.

Let $S_k=Y_1+\dots+Y_k$, and $S_k^+=\max(0,S_k)$

I want to know a close form of $\mathbb{E}(S_k^+)$

share|cite|improve this question
up vote 1 down vote accepted

Let $X$ denote a discrete random variable with finite mean $\mu$. From $$E[X-\mu] = 0 = \sum_i (u_i - \mu) p_X(u_i) = \sum_{i\colon u_i > \mu} (u_i - \mu) p_X(u_i) + \sum_{i\colon u_i < \mu} (u_i - \mu) p_X(u_i),$$ we get that $$\sum_{i\colon u_i > \mu} (u_i - \mu)p_X(u_i) = \sum_{i\colon u_i < \mu} (\mu - u_i) p_X(u_i) $$ and so if $Y = \max\{0, X-\mu\}$, we have that $$\begin{align*} E[Y] &= \sum_i \max\{0, u_i - \mu\} p_X(u_i)\\ &= \sum_{i\colon u_i > \mu} (u_i - \mu)p_X(u_i)\\ &= \sum_{i\colon u_i < \mu} (\mu - u_i) p_X(u_i). \end{align*}$$ In the question under consideration, $X = \sum_{j=1}^k X_j$ is a negative binomial random variable with parameters $(k,p)$ and mean $\frac{k}{p}$. Its pmf is given by $$p_X(n) = \binom{n - 1}{k-1} p^k (1 - p)^{n-k}, ~~ n = k, k+1, k+2, \ldots $$ Hence, $$\begin{align*} E[S_k^+] &= E\left[\max\left\{0, X-\frac{k}{p}\right\}\right]\\ &= \sum_{n = \lceil \frac{k}{p}\rceil}^\infty \left(n - \frac{k}{p}\right)\binom{n - 1}{k-1} p^k (1 - p)^{n-k}\\ &= \sum_{n = k}^{\lfloor \frac{k}{p}\rfloor} \left(\frac{k}{p}- n \right)\binom{n - 1}{k-1} p^k (1 - p)^{n-k}. \end{align*}$$ If $k/p$ is an integer, one of the terms in each sum is $0$. The first form of the answer has already been given by @Alen but the second might be easier to evaluate numerically since it is a finite sum.

Neither form gives the floor of the expected value as has been alleged by OP Fan Zhang in comments on the main question.

share|cite|improve this answer

Since $\sum\limits_{i = 1}^k {{X_i}} $ has negative binomial distribution with parameters $k$ and $1-p$, we have $$ \mathbb{E}\left[ {S_k^ + } \right] = \sum\limits_{j = \left\lceil {\frac{k} {p}} \right\rceil }^\infty {\left( {j - \frac{k} {p}} \right)\left( \begin{gathered} j + k - 1 \\ j \\ \end{gathered} \right){p^k}{{\left( {1 - p} \right)}^j}} $$

share|cite|improve this answer
This disregards jensen's inequality, and you can calculate it easier/to a number. – gnometorule Jan 8 '12 at 14:57
...or not to the 'easier' part. My bad. It still seems to disregard jensen's inequality. – gnometorule Jan 8 '12 at 15:09
@gnometorule: What are you talking about? – Did Jan 8 '12 at 15:14
Alen: Added a link. – Did Jan 8 '12 at 15:17
Is $\mathbb{E}\left[ {S_k^ + } \right] = \mathbb{E}[S_k^+]$? – Fan Zhang Jan 8 '12 at 16:22

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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