[gmx-users] autocorrelation function using g_analyze

[Mark Abraham]

Sung Hyun Park wrote:
> Hello,
> 
> Does anyone know exactly how g_analyze -ac calculates autocorrelation function?
> My input file, dist.xvg, contains a distance d(t) between two specific
> atoms of my molecule in time t. (hence d(t)>0)
> I calculated the autocorrelation function using the following command:
> 
> g_analyze -f dist.xvg -ac corr-dist.xvg
> 
> What I don't understand is that the correlation function C(k) in the
> output file, corr-dis.xvg, includes some negative values and even goes
> down to almost -0.5!
> I thought by definition the autocorrelation function C(t) is
> calculated as following:
> 
> C(k) = <d(t)*d(t+k)>
> 
> where C(k) should be always positive because d(t), distance, is always
> positive in this case.
> So does anybody know precisely what formula "g_analyze -ac" uses to
> calculate the autocorrelation function?

They're described in the manual in one of the appendices.

Mark