I'm reading a paper which has the following description;

Say we have a time series of correlated sequential observations of the random variable $X$ denoted $\{x_n\}_{n=1}^N$ from a stationary, time reversible stochastic process.

Expectation of $X$ is

$\hat{X} = \dfrac{1}{N}\sum_{n=1}^Nx_n$

and statistical uncertainty is

$\delta^2\hat{X} \equiv \langle(\hat{X}-\langle\hat{X}\rangle)^2\rangle = \langle\hat{X}^2\rangle - \langle \hat{X} \rangle^2$

Which can be written as

$\delta^2 \hat{X} \equiv \dfrac{1}{N^2}\sum_{n,n' = 1}^N \bigg[ \langle x_n x_{n'} \rangle - \langle x_n \rangle \langle x_{n'} \rangle \bigg]$

Focussing on the summation, what is meant by the $[n,n'=1]$ notation? I'm suspect it's shorthand for

$\delta^2 \hat{X} \equiv \dfrac{1}{N^2}\sum_n^N\sum_{n'}^N \bigg[ \langle x_n x_{n'} \rangle - \langle x_n \rangle \langle x_{n'} \rangle \bigg]$

But I don't know why you wouldn't just write that.

EDIT1 Moreover, in this context, it's not clear to me why $\langle x_n \rangle$ and $\langle x_{n'} \rangle$ are not simply both equal to $\hat{X}$, assuming we read $\langle x_n \rangle$ as "the expected value of a member of the set generated by the random variable $X$". Or perhaps we should read $\langle x_n \rangle$ as the series from $n$ to $N$, such that as we increment $n$ the series of values becomes shorter (which I think makes intuitive sense)?

EDIT 2 Sorry - ignore all that, $\langle x_n \rangle$ is explicitly defined as the expected value of $x_n$ over repreated independent trials (i.e. Generating 100 independent series and determinig the average value of the $n$th member).

  • 1
    $\begingroup$ I think you misplaced your $\dfrac 1 {N^2}$ $\endgroup$ – GFauxPas Jan 19 '15 at 18:16

It's a double summation.


Of course the given notation is more convenient.


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.