Version 2.0.4

This is a somewhat significant release.  After extensive discussions with (and significant code contributions from) Michael Shirts (and through him John Chodera), I've revamped the bootstrapping uncertainty analysis again.  Prior versions of the code computed the uncertainty in the free energy by transforming the uncertainty in the probability, assuming those values were Gaussian-ly distributed.  The current code does a better job, aligning the PMFs themselves by insisting that their Boltzmann-averaged free energies are the same.  This choice is arbitrary but reasonable.

At the same time, I've found some significant flaws in the 2D version of bootstrapping, which I believe are separate from these problems, and were masked by the other flaws.  Rather, they seem to spring from the assumption that it's ok to collapse a 2D distribution into a 1D cumulant.  For this release, I've simply removed the ability to run 2D bootstrapping, since I don't have a version I trust.  I hope to fix this soon.

Download here

June 20, 2010