I used the gmx msd function to calculate the mean square displacement and estimate the diffusion constant of atoms in my simulations. The *.xvg file reports the diffusion constant and also gives a range (+/- 0.0003, for example).
Is this range the standard deviation? Or is it the standard error? If it's the standard error, what is the confidence level (90%, 95%, 99%)? Is there a way to toggle the confidence level?
I looked in the most recent manual but couldn't manage to find this information. Perhaps, I'm looking in the wrong spot.
http://manual.gromacs.org/current/onlinehelp/gmx-msd.html
"An error estimate given, which is the difference of the diffusion coefficients obtained from fits over the two halves of the fit interval."
You can see how everything is calculated in the code, src/gromacs/gmxana/gmx_msd.cpp in the function do_coor().
Justin Lemkul Thank you for highlighting that.
I think I'll run the MSD analysis on several blocks of the trajectory then so I can average their diffusion constants and calculate my own standard error with a 95% confidence interval.
Statistical sampling can be very difficult, depending on how short each interval is. Depending on what species is diffusing, you may be better off running multiple, independent simulations and computing a value for each. The small error listed above indicates a pretty well-converged simulation, but I figured I would mention it.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics