### Regression test of muVT ensemble using flat files
### System: GenericMuVT
## Reading first result
## Reading second result
## Validating ensemble
###############################################################################
# WARNING: Support for muVT ensemble is an experimental feature under current #
#          development. You can help us to improve it by reporting errors     #
#          at https://github.com/shirtsgroup/physical_validation              #
#          Thank you!                                                         #
###############################################################################
Analytical slope of ln(P_2(U - \sum mu*N, N)/P_1(U - \sum mu*N, N)): 0.00657226, 0.44362755
Trajectory 1 equilibration: First 87 frames (8.7% of trajectory) discarded for burn-in.
Trajectory 1 decorrelation: 76 frames (8.3% of equilibrated trajectory) discarded for decorrelation.
Trajectory 1 tails (cut = 0.10%): 3 frames (0.36% of equilibrated and decorrelated trajectory) were cut
After equilibration, decorrelation and tail pruning, 83.42% (835 frames) of original Trajectory 1 remain.
Trajectory 2 equilibration: First 29 frames (2.9% of trajectory) discarded for burn-in.
Trajectory 2 decorrelation: 210 frames (21.6% of equilibrated trajectory) discarded for decorrelation.
Trajectory 2 tails (cut = 0.10%): 4 frames (0.52% of equilibrated and decorrelated trajectory) were cut
After equilibration, decorrelation and tail pruning, 75.72% (758 frames) of original Trajectory 2 remain.
Overlap is 2.0% of trajectory 1 and 2.8% of trajectory 2.
A rule of thumb states that a good overlap can be expected when choosing state
points separated by about 2 standard deviations.
For the current trajectories, dT = 5.0, and dmu = 0.5,
with standard deviations sig1 = [51.9, 8], and sig2 = [74.5, 1e+01].
According to the rule of thumb, given point 1, the estimate is dT = 28.8, dmu = 0.7, and
                                given point 2, the estimate is dT = 20.8, dmu = 0.4.
Rule of thumb estimates that (dT,dmu) = (24.8,0.5) would be optimal (currently, (dT,dmu) = (5.0,0.5))
Computing log of partition functions using pymbar.BAR...
Using -39.90363 for log of partition functions as computed from BAR.
Uncertainty in quantity is 0.20954.
Assuming this is negligible compared to sampling error at individual points.
Computing the maximum likelihood parameters
==================================================
Maximum Likelihood Analysis (analytical error)
==================================================
Free energy
    -16.40757 +/- 10.03476
Estimated slope                  |  True slope
    0.003281  +/- 0.009705       |  0.006572 
    (0.34 quantiles from true slope)
    0.181417  +/- 0.122671       |  0.443628 
    (2.14 quantiles from true slope)
Estimated dT                     |  True dT
    2.5    +/- 7.4               |  5.0   
Estimated dmu                    |  True estimated dmu
    -0.5   +/- 0.3               |  -1.1  
==================================================
Computing bootstrapped maximum likelihood parameters
==================================================
Maximum Likelihood Analysis (bootstrapped error)
==================================================
Free energy
    -16.40757 +/- 9.90576
Estimated slope                  |  True slope
    0.003281  +/- 0.026711       |  0.006572 
    (0.12 quantiles from true slope)
    0.181417  +/- 0.056632       |  0.443628 
    (4.63 quantiles from true slope)
Estimated dT                     |  True dT
    2.5    +/- 20.3              |  5.0   
Estimated dmu                    |  True estimated dmu
    -0.5   +/- 0.1               |  -1.1  
==================================================
Calculated slope is (0.1, 4.6) quantiles from the true slope
