[gmx-users] converging value

chris.neale at utoronto.ca chris.neale at utoronto.ca
Sun Apr 10 15:06:49 CEST 2011


You need to look at the dependence of the statistical value (in this  
case the average pressure) with (a) increasing equilibration time that  
you discard prior to collecting pressure data, and (b) increasing  
production simulation time that you include in your average.

Create the first plot, and discard any data in the early part of the  
simulation for which there appears to be a drift (the discarded data  
will not be used in the second plot). Then create the second plot,  
where you should see large oscillations about an average value and  
these oscillations should get smaller as the simulation time is  
increased. You can gauge the remaining error by block averaging or  
(much better) comparing with another simulation that you ran. There  
are also other statistical methods available, but I prefer the  
simplicity of these.

If your question is simply that you did these things but the  
oscillations remain large and you want to know how to get a more  
precise value, then the answer is simply to simulate longer.

Chris.

-- original message --

Dear All

How can I determine the converged value of a simulation?
Because the pressure has big oscilations around it's converging value but it
is difficult to determine that value.

can everybody guide me?
Thanks in advance
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.gromacs.org/pipermail/gm




More information about the gromacs.org_gmx-users mailing list