<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
Hi,<br>
<br>
Next time please file a bugzilla.<br>
I filed one now:<br>
<a class="moz-txt-link-freetext" href="http://redmine.gromacs.org/issues/870">http://redmine.gromacs.org/issues/870</a><br>
I posted a fix as well (there were more things to fix).<br>
Please try it and report back on redmine.<br>
<br>
Cheers,<br>
<br>
Berk<br>
<br>
On 01/25/2012 06:40 PM, Ivan Gushchin wrote:
<blockquote
cite="mid:CABxJBLY54cZmDcD_FCLAz_3xrnkApKXSc7EOFEmPzqqWBBAdMA@mail.gmail.com"
type="cite">
<meta http-equiv="Content-Type" content="text/html;
charset=ISO-8859-1">
Hi,<br>
<br>
There seems to be a bug in g_anaeig (4.5.3) related to the
determination of the extreme (structural) projections of the
trajectory on the eigenvector, when more than one eigenvector is
analyzed. When I use different numbers of the eigenvectors for
analysis of the extreme projections, the generated fluctuations
for the same eigenvector differ in amplitude. The problem results,
as I understand, from the fact that the maximum and minimum
projections used for the generation of the output files
(gmx_anaeig.c lines 663-665) come from the last analyzed
eigenvector (defined by -last).<br>
<br>
lines 624-636, determination of the extreme projections:<br>
<br>
for(v=0; v<noutvec_extr; v++) {<br>
for(i=0; i<nframes; i++) {<br>
if (inprod[v][i]<inprod[v][imin[v]])<br>
imin[v]=i;<br>
if (inprod[v][i]>inprod[v][imax[v]])<br>
imax[v]=i;<br>
}<br>
<b>min=inprod[v][imin[v]];<br>
max=inprod[v][imax[v]];<br>
</b>fprintf(stderr,"%7d %10.6f %10.1f %10.6f %10.1f\n",<br>
eignr[outvec[v]]+1,<br>
min,inprod[noutvec][imin[v]],max,inprod[noutvec][imax[v]]); <br>
}<br>
<br>
lines 663-665, generation of the output coordinates:<br>
<br>
xread[index[i]][d] = <br>
(xav[i][d] + (<b>min</b>*(nextr-frame-1)+<b>max</b>*frame)/(nextr-1)<br>
*eigvec[outvec[v]][i][d]/sqrtm[i]);<br>
<br>
As it can be seen, for all the eigenvectors the values of the
extreme projections used are the same, which is incorrect, as the
vectors are normalized but fluctuations along different vectors do
differ in amplitude. It can be healed by the following
modification:<br>
<br>
xread[index[i]][d] = <br>
(xav[i][d] + (<b>inprod[v][imin[v]]</b>*(nextr-frame-1)+<b>inprod[v][imax[v]]</b>*frame)/(nextr-1)<br>
*eigvec[outvec[v]][i][d]/sqrtm[i]);<br>
<br>
Could someone check if this is correct? <br>
<br>
Please excuse me if I post to the wrong mailing list.<br>
<br>
All the best,<br>
Ivan.<br>
<br>
<fieldset class="mimeAttachmentHeader"></fieldset>
<br>
</blockquote>
<br>
</body>
</html>