On Thu, Aug 18, 2011 at 4:24 PM, James T Metz
<address@hidden> wrote:
Gabor,
Thank
you for your response.
I
have already tried the using cmdscale and smacof within R.
My
input distance matrix contains many "long" distances. The
output 2D coordinates from cmdscale
do not reproduce my input distance
matrix very well. I am getting
distance errors on the order of the dimensions
of the 2D map! Hence my points
might as well be anywhere!
I
am getting similar behavior with smacof.
I
have also tried various weighting schemes with little success so
far.
I
am trying Igraph since there might be some advantage to having my
nodes being layed out in 2D and not
overlapping. This approach might work
for some cases, but may not be useful
in general.
Regards,
Jim
Metz
James T. Metz, Ph.D.
Senior Scientist III Chemistry
GPRD R4DG AP52 - 01D10P
Abbott Laboratories
100 Abbott Park Road
Abbott Park, IL 60064 - 6100
U.S.A.
Office (847) 936 - 0441
FAX (847) 938 - 2756
address@hidden
This communication may contain information that is legally privileged,
confidential, or exempt from disclosure. If you are not the intended
recipient, please note that any dissemination, distribution, use, or copying
of this communication is strictly prohibited. Anyone who receives
this message in error should notify the sender immediately by telephone
(847 - 936 - 0441) or return email and delete it from his or her computer.
From:
| Gábor Csárdi <address@hidden>
|
To:
| Help for igraph users <address@hidden>
|
Cc:
| |
Date:
| 08/18/2011 05:15 AM
|
Subject:
| Re: [igraph] best method of preserving
network interactions in a 2D map?
|
Sent by:
| address@hidden |
On Sun, Aug 14, 2011 at 9:35 PM, James T Metz <address@hidden>
wrote:
>
> Igraph List,
>
> Suppose I have a set of interactions e.g.,
similarities or distances
> where distance = 1 - similarity
> for a set of nodes. I can read in those interactions, generate
a network
> and output the 2D coordinates.
>
> Suppose I would like to preserve the input
similarities (or
> distances) as much as possible so that
> if I computed the Euclidean distances from the resulting network 2D
> coordinates, I would get an excellent
> agreement with the original input distances (similar to multidimensional
> scaling).
>
> Is this possible using the igraph package?
If so, what functions
> and options should be used.
If you use R, then you don't need igraph for this, see the cmdscale() function.
Gabor
[...]
--
Gabor Csardi <address@hidden> MTA KFKI RMKI