# Sum of Bernoulli variables with different success probabilities [duplicate]

This question already has an answer here:

Let $x_i$ be independent Bernoulli random variables with success probabilities $p_i$. That is, $x_i=1$ with probability $p_i$ and $x_i=0$ with probability $1-p_i$.

Is there a closed expression or an approximate formula for the distribution of the sum $\sum_i x_i$?

-

## marked as duplicate by whuber♦Apr 15 at 16:57

If the $p_i$ are very small, you can use Poisson approximation. Let $X_i\sim \mbox{Be}(p_i)$ be independent and let $Y\sim\mbox{Po}(\lambda)$ with $\lambda=\sum_{i=1}^np_i$. In a classic paper by Hodges and Le Cam it is shown that $|\mbox{P}(\sum_{i=1}^n X_i\leq x)-\mbox{P}(Y\leq x)|=3\cdot (\max_{1\leq i\leq n}p_i)^{1/3}.$ If the $p_i$ are all close to 0, this difference is small. –  MånsT Apr 15 at 13:15
In addition to the duplicate, solutions appear at stats.stackexchange.com/questions/41247 (computational methods) and stats.stackexchange.com/questions/5347 (approximations for large numbers of variables). –  whuber Apr 15 at 16:59

Yes, in fact, the distribution is known as the Poisson binomial distribution, which is a generalization of the binomial distribution. The distribution's mean and variance are intuitive and are given by

\begin{align} E\left[\sum_i x_i\right] &= \sum_i E[x_i] = \sum_i p_i\\ V\left[\sum_i x_i\right] &= \sum_i V[x_i] = \sum_i p_i(1-p_i). \end{align}

The expectation is straightforward because it is a linear operator. The variance is also straightforward because of the independence assumption.

-

I'm not aware of a closed formula to exist. If n becomes relevant you can apply Central Theorem Limit so approximating the sum distribution with a normal distribution having mean the sum of p_i and variance the sum of p_i * ( 1 - p_i).

-
The CLT often fails (that is, does not even apply) in this circumstance unless the $p_i$ remain away from $0$ and $1$. –  whuber Apr 15 at 17:01