Suppose I take $n$ measurements of a discrete random variable in an experiment, and from that get a mean:

$ \mu = \frac{1}{n}\sum{x_i}$

I take this to be the mean in a Poisson distribution. Is the uncertainty in that mean the standard error:

$s_n = (\frac{\mu}{n})^\frac{1}{2}$ ?


Yes if the sample is independent and identically distributed, this is an appropriate expression for the standard error of the mean. It is a direct result of both 1) the law of total variance which states that $\mbox{var}(X+Y) = \mbox{var}(X) + \mbox{var}(Y) + 2 \mbox{cov} (X, Y)$ (where the last term is 0 if $X,Y$ are independent) and 2) appealing to the closure of Poisson distributions under a sum: so that $\sum_{i=1}^n X_i$ has a Poisson($n\mu$) distribution with variance $n\mu$ and the sample mean $\bar{X}$ is $$\mbox{var}(\frac{1}{n} \sum X) = \frac{1}{n^2}n\mu=\frac{\mu}{n}$$

And the variance of the sum is transformed to a standard deviation through the square root.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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