# Re: followup question on error calculation in ABF

From: Jerome Henin (jhenin_at_cmm.chem.upenn.edu)
Date: Thu Oct 08 2009 - 04:04:51 CDT

Hi everyone,

To evaluate the correlation length, you need the time series of F_xi
(and some skill, and possibly some luck).

In the old implementation (NAMD 2.6), the time series is obtained
using the keyword writeFxiFreq, which I realize is undocumented, as
discussed a couple of years ago on this list:
http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l/6712.html
(and then the code was obsoleted, so I never got around to fixing the
documentation)

In NAMD 2.7b1, that's done at the colvar level; the keyword is
outputSystemForce, and the data ends up in the colvar trajectory file.
http://www.ks.uiuc.edu/Research/namd/2.7b1/ug/node48.html

Cheers,
Jerome

2009/10/7 Ramya Gamini <ramyabhargavi_at_gmail.com>:
>
> Sorry! Giovanni, but I cannot help you with this part of your query.  From
> what I understood from your previous query, I only thought you weren't able
> to compute the factor (1+2k) as you did not know where to obtain the
> instantaneous forces from.
> I haven't performed an error analysis for ABF simulation myself but I am
> sure Jerome or other ABF experts could best explain on how to
> evaluate (1+2k)
> best,
> Ramya
> On Wed, Oct 7, 2009 at 3:35 PM, Giovanni Bellesia <gbellesia_at_chem.ucsb.edu>
> wrote:
>>
>> Hi Ramya,
>>
>> the distFile is exactly what I'm using to calculate (for each bin) the
>> standard deviation via a Gaussian fit
>> of the distributions.
>> I was just wondering how I should calculate the factor (1+2k) which is
>> related to the correlation length of the instantaneous force.
>>
>> Giovanni
>>
>> Hi Giovanni,
>>
>> As I understand .. .. ..the 'distFile' key word in the config file, will
>> allow you generate the '.dist' file that includes instantaneous components
>> of the force.
>> For each bin, the number of samples for instantaneous forces ranging from
>> -fMax to fMax will be generated. Default of fMax being 60 kcal/mol/A
>>
>> Like these terms are defined in the namd -user guide under ABF section.
>> distFile   Output file containing force distributions
>> Acceptable Values: Unix filename
>> Default Value: none
>> Description: Output file containing a distribution of the instantaneous
>> components of the force, , for every bin comprised between xiMin and xiMax.
>> This is useful for performing error analysis for the resulting free energy
>> profile
>>
>> fMax   Half-width of the force histograms
>> Acceptable Values: positive decimal number, in kcal/mol/Å
>> Default Value: 60.0
>> Description: When force distributions are written in distFile, the
>> histogram collects  values ranging from fMax to fMax.
>>
>>
>> Ramya
>>
>> On Wed, Oct 7, 2009 at 1:50 PM, Giovanni Bellesia
>> <gbellesia_at_chem.ucsb.edu> wrote:
>>>
>>> Hi all,
>>> I am trying to use formula (22) in JCP 121(7), 2004 to calculate the
>>> error in my ABF data.
>>> From the simulations, I have the distributions of the instantaneous force
>>> for every bin.
>>> Therefore, I am calculating sigma(F_{e,i}) fitting these distributions
>>> with a Gaussian.
>>>
>>> My problem is: I understand the definition of "sampling ratio" (1+2k) but
>>> I don't know how to
>>> calculate it since I don't have the values of the instantaneous force.
>>> I think I'm missing something very simple here ... :o)
>>>
>>> Giovanni
>>>
>>> --
>>> ------------------------------------------
>>> Giovanni Bellesia
>>> Department of Chemistry and Biochemistry
>>> University of California, Santa Barbara
>>> Santa Barbara, California 93106-9510, USA.
>>> Email: gbellesia_at_chem.ucsb.edu
>>> Phone: 805 893 2767
>>> Web: http://www.chem.ucsb.edu/~gbellesia/
>>> ------------------------------------------
>>>
>>
>>
>>
>> --
>> ------------------------------------------
>> Giovanni Bellesia
>> Department of Chemistry and Biochemistry
>> University of California, Santa Barbara
>> Santa Barbara, California 93106-9510, USA.
>> Email: gbellesia_at_chem.ucsb.edu
>> Phone: 805 893 2767
>> Web: http://www.chem.ucsb.edu/~gbellesia/
>> ------------------------------------------
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:21 CST