Standard error on AWH free energy estimate?

Is it possible to get a standard error associated with a free energy estimate with gmx awh similar to what is given by gmx bar? I tried using gmx bar on the dhdl.xvg file and it just spits out a segmentation fault.

Hello,

I’m tagging Berk (@hess) here for the general question, but the tool shouldn’t segfault in any case.

Can you please open a bug report on Gitlab (Issues · GROMACS / GROMACS · GitLab) and tag me (@acmnpv) there? Please also add some example files that reproduce the error.

Cheers

Paul

No, there is no error estimate for a single AWH run. Since the AWH sampling is time dependent one can not do simple error estimates. A rough error estimate can be obtained by comparing free-energies at different times. But for proper error estimates we recommend running a few indepedent AWH runs.

The segfault occurs with both AWH and Expanded Ensemble-generated dhdl files. Perhaps it’s due to the extra thermodynamic state column?

We will need the input files to check if there is a bug in the tool.

Can you please upload them somewhere?

Cheers

Paul

| carruthj
May 30 |

  • | - |

The segfault occurs with both AWH and Expanded Ensemble-generated dhdl files. Perhaps it’s due to the extra thermodynamic state column?

awhdhdl.xvg (7.2 MB)
eedhdl.xvg (7.1 MB)

You should not analyze AWH with gmx bar as the data has not been collected at equilibrium.

Also not all input required for gmx bar is in the output file. A too late check on the input causes a segmentation fault in gmx bar.