[gmx-users] block averaging in g_energy vs. g_analyze

[Hanne Antila]

Dear gurus,

I'm trying to calculate some statistics for the coulomb-sr and
coulomb-recip energies in my system (200 ns simulation). I Suppose the
underlying algorithms in g_energy and g_analyze for making the error
estimate are pretty much the same, but in g_energy the sampling is
more accurate which can cause some differences. However, I'm confused
about two points

1. What can be the cause of nan as g_energy output rmsd when the mean
is ok (and also g_analyze gives ok rmsd)?

2. g_analyze gives me an error regarding set 1. I think this error
implies not-long-enough sampling (longer correlation than run time),
but can there be another cause for this? In this case the correlation
would be very, very long.

I have enclosed the output from g_energy and g_analyze.

All help will be appreciated!

Yours,

Hanne


g_energy:

the last energy frame read 10002 time 200000.000

Statistics over 99984548 steps [ 0.0000 through 200000.0000 ps ], 2 data sets
All statistics are over 16860000 points

Energy                      Average   Err.Est.       RMSD  Tot-Drift
-------------------------------------------------------------------------------
Coulomb (SR)             -1.98611e+06         27        nan    133.058  (kJ/mol)
Coul. recip.                -470752                 5.7        nan
-18.7256  (kJ/mol)

g_analyze:

Read 2 sets of 10003 points, dt = 19.996

                                      std. dev.    relative deviation of
                       standard       ---------   cumulants from those of
set      average       deviation      sqrt(n-1)   a Gaussian distribition
                                                      cum. 3   cum. 4
SS1  -1.986081e+06   2.158308e+03   2.158093e+01       0.003    0.006
SS2  -4.707511e+05   9.907673e+01   9.906682e-01       0.005    0.014

a fitted parameter is negative
invalid fit:  e.e. 22.6293  a 1.04171  tau1 14.54  tau2 99.5578
Will fix tau2 at the total time: 200000
Set   1:  err.est. 76.0536  a 0.999443  tau1 12.7992  tau2 200000
Set   2:  err.est. 6.73835  a 0.926159  tau1 13.7527  tau2 6092.31

gcq#135: "Check Your Input" (D. Van Der Spoel)