[gmx-users] estimation of error using ACF.

[Alan Wilter Sousa da Silva]

Thank you very much Dr. Spoel.

	I played with g_lie and find two small errors.

	Changing g_lie.c, I did (commentaries):

static real lie_lj=0,lie_qq=0,fac_lj=0.161,fac_qq=0.5; /* 0.181 by 0.161*/

and

    { "-Cqq",  FALSE, etREAL, {&fac_qq}, /* lie by fac */

	I also noted, but I wasn't able to fix, that at lie.xvg output,
last line has the value 0 (zero) for column Time and repeats the the last
value for DGbind.  So when using g_analise I got error with option -ee.
Removing that last line, g_analise works fine and then I got a err.est,
ACF, as well the right value for average, which is a little bit different
from what g_lie gives, certainly due to the extra datum.

Do you think that such errors in code of g_lie.c should be corrected,
supposed I did the right thing?  And, if you don't mind, since I'm not
familiar to CVS, of course if you wish, would you make that corrections in
CVS version?

I hope to be contributing a little bit to Gromacs, although I'll never get
to pay back all what GMX has been doing for me.

Sincerely,

On 11 Feb 2003, David wrote:

> there is a special program g_lie to do this kind of things. Double check
> your output, though, it is not very well tested. Error estimates are
> best done by g_analyze (on your .xvg file), although knowledge of the
> energy autocorellation can be useful.

-----------------------
Alan Wilter S. da Silva
-----------------------
 Laboratório de Física Biológica
  Instituto de Biofísica Carlos Chagas Filho
   Universidade do Brasil/UFRJ
    Rio de Janeiro, Brasil