[gmx-users] Re: autocorrelation function using g_analyze

[Mark Abraham]

Justin A. Lemkul wrote:
> 
> 
> Sung Hyun Park wrote:
>>> They're described in the manual in one of the appendices.
>>>
>>> Mark
>>
>> Dear Mark,
>>
>> Thank you so much for your reply.
>> I just checked the manual again and looked through the appendices,
>> but could not find the exact formula used by gromacs to calculate the
>> autocorrelation function in g_analyze.
>> I'm sorry to bother you again but could you tell me more specifically
>> where to look in the appendices?
>> I also checked the special topics chapter in the manual but it is
>> still not clear to me which equation is being used by gromacs.
>>
> 
> I would suggest section 8.5 "Correlation functions" to get the overview 
> of these calculations, and any special forms that might be used be 
> certain programs.

Indeed. The key bit of information not explicitly stated in 8.5, but 
present in g_analyze -h, is that normalization is on by default. The 
correlation function is only dimensionless if it has been normalized 
(subtract mean from all data points, divide by standard deviation). A 
normalized correlation function can readily be negative - consider the 
correlation of two fully-out-of-phase sinusoids.

Mark