search

alchemistry / alchemical-best-practices

Best practice document for alchemical free energy calculations going to livecoms journal
Creative Commons Attribution 4.0 International
63 stars 18 forks source link

[review] accuracy increase for longer simulations? #107

Closed ppxasjsm closed 4 years ago

ppxasjsm commented 4 years ago

Is this the case?

Can the authors give some rules of thumb how fast the accuracy increases with simulation length? Is it always as simple as /sqrt(N)?

mrshirts commented 4 years ago

It's basically as simple as 1/sqrt(N), where N is the number of uncorrelated samples. In most cases, the uncertainty of free energy calculations does behave as normally distributed, and thus the 1/sqrt(N) calculation applies. This needs a bit more thought, though, to make it clear why it's the case. As a rule of thumb, this is fine, though.

davidlmobley commented 4 years ago

Was this a comment on a particular line number? Could help with edits along these lines.

ppxasjsm commented 4 years ago

I would imagine this is referring to section 7.2.5

davidlmobley commented 4 years ago

Added this at the end of section 7.2.5 on Overleaf. Please review and close this issue if you're happy with the changes.

mrshirts commented 4 years ago

Tweaked a bit more, looks good to me now.

ppxasjsm commented 4 years ago

Great thanks! Will take a look in my big sweep.