Re: [AMBER] ncorr in rotdif command seems redundant to me. Please help.

From: David A Case <>
Date: Wed, 4 Nov 2015 07:16:37 -0500

On Wed, Nov 04, 2015, Jose Borreguero wrote:
> I was also wondering if there was any kind of "ensemble average" when
> calculating the correlation function by averaging over different trajectory
> chunks. For instance, if ncorr=1000 then calculating C(t) over frames
> [0,1000], [1000,2000], [2000,3000],... and then doing the average of these
> C(t)'s. From what I gather this average is not an option right now, correct?

It's not possible, other than by running a bunch of cpptraj runs with
different frames selected by the trajin command, then writing a little script
to average the C(t) values you would get.

Beyond that, this would be a bad idea. You would lose any information
about what happened (say) between frame 500 and frame 1500, since none
of the individual runs would see that combination. If you only use 1000
frames, your C(t) is probably good only to about 50 frames. If you use
10,000 frames, you can probably get a fairly good time correlation out to
500 frames. etc.

Having said that, give what you suggested a try, and see what happens! That
is the real way to learn. There are places where averaging C(t) curves is a
real win, but I don't think that is true in your case. Seeing for yourself
what is going on with different methods of analysis is far better than trying
to parse my (often terse and sometimes obtuse) emails.


AMBER mailing list
Received on Wed Nov 04 2015 - 04:30:03 PST
Custom Search