Ring Cohesion
    

progress in methods: completed
(1) Pajek (Andrej Mrvar) added an automated procedure for us (see below)
(2) have generated all the 3-cycles for biotech 1- and 2- mode (360 of them; 
    220 nonisomorphic 3-cycles identified by Joseph Wehbe)
(3) no pair of the 220 are isomorphic (software needed for checking)
(4) drw program created the fragment files in pajek *.paj
(5) these data used to do fragment frequencies
(6) and for testing with Pajek (one fragment took 2-3 seconds, should be about 10 minutes for all 220.
(7) project done for 3-cycles, probably feasiblealso for 4-cycles.


From andrej.mrvar@Uni-Lj.si Fri Aug  8 15:25:45 2003
Date: Sun, 08 Aug 2004 22:10:10 +0200
From: Andrej Mrvar 
To: Douglas R. White 
Subject: Try this

Hello!

I made a modification in Pajek (download) and prepared a file for running  
iterative fragment searching.
Some instructions:

***************************************************************************
1. Run Pajek

1. Select appropriate options to read genealogy as Ore graph:

    Options/ReadWrite/ GEDCOM p-graph - not marked
    Options/ReadWrite/ Ore 1-male, 2-female links - marked

2. Run logfile allfragmen.log that is attached, using:
    File/RepeatSession and select allfragmen.log

3. On the first question you must select Pajek project file with fragments,
    that you prepared and I made some modifications (pjk0001.paj).

4. On the second question select network with genealogy stored in GEDCOM
file
    (like p-tur.ged)

5. After some running you will get a lot of vectors in Pajek.
    All vectors are also written to text file VECTORS.VEC in the directory
where you selected the
    GEDCOM file (the first four columns are INDI, BIRTH, MARR, DEATH
    and than 227 vectors for your fragments.

There is one more request: since in all your fragments the first person
is male (triangle) also in the gedcom file the first person must
be male, otherwise you will get only zeros in all vectors.
If the first person in GEDCOM file is a woman, run  allfragwom.log
instead of allfragmen.log.

I hope the explanaion is clear enough, and you will get the results you
need.
In the attached file you will find
- new Pajek.exe
- paj0001.paj - your list of fragments
- allfragmen and allfragwom: sessions for several fragment searching

Best,
Andrej


Hello Again!

> One item: if I click ( ) induced subgraph I get everything inside a cycle
> as well as a cycle, right. and If I leave it unclicked I can get just a
> cycle fragment with nothing inside it?

It is just the opposite. Suppose that we are searching for cycles:

- if induced is checked no additional lines between vertices belonging to
  the cycle found are allowed (if there are some additional lines, Pajek
says this is no matching)

- if induced is unchecked, and there are some additional lines between
vertices in the
  cycle, this is considered as a matching pattern

To demonstrate it, I attached two files
- fragment frag.net (first network)
- sample network sampl.net (second network)

If you run Fragment searching with option Induced checked, no fragments will
be found,
but if you uncheck this option, 1 fragment will be found (there is
additional line from vertex 1 to vertex 6).
So in general, if induced is not checked, more fragments can be found.

I made some modifications of fragment searching, therefore I attached Pajek
as well.

Best,

Andrej





Date: Thu, 23 Oct 2003 22:34:59 +0200
From: Andrej Mrvar 


Hello!

...Any luck on processing the fragments?

Searching fragments where matching in types of vertices must be performed:
In this case you must provide partition of vertices in original network and
partitions of vertices in fragments and demand matching of classes using
Nets/Fragments/Options/Labeled
I prepared files 1212.clu and pjk_bio1.paj
(lines following empty lines are description of edges, they are skipped by
Pajek,
they can be deleted as well, but you need some time to do that)..

I prepared new log file (File/Repeat Session) when asked for files answer:

1. pjk_bio.paj
2. pjk_bio1.paj
3. 1212.net
4. 1212.clu - classifies nodes by type of organization, 1-6

And all the rest will be done (result in vectors.vec).
All network files, macro and reculting vectors.vec are attached

Best,
Andrej



----------------------------------------

Hello!

>One question: since these are all 3-cycles, wouldnt you expect
>the sum of numbers in the vector columns to add up to multiples of 3 and
>to equal 3*frequency as computed in the *.rep file?

The numbers you get in vectors.vec are 1, 2, 3, 4, 5, or 6,
and these are in your example the types (colors) of vertices that are in the
fragment
and not the counts, as you think.

If would like to get the counts (how many times each vertex belongs
to specified fragment, you must use result of fragment searching that
is stored in a partition.

I attached the modified log file (bio_frag1.log) that saves that result in
vectors1.vec as well (except the first column that is actually the input
partition - numbers 1, 2, 3, 4, 5, or 6).
vectors1.vec is also attached.

Andrej


    

    Search

    

    Topic outline
    1
    Università di Modena e Reggio Emilia
    
Resource Guido Fioretti
    
Resource Andrea Ginzburg
    
Resource David A. Lane
    
Resource Federica Rossi
    
Resource Margherita Russo
    
Resource Roberto Serra
    
Resource Marco Villani
    


    



    2
    CNRS UMR 7041, Archéologies et Sciences de l' Antiquité, Paris
    
Resource Sander van der Leeuw
    


    



    3
    CNRS UMR 8504 - Géographie-cités èquipe, Paris
    
Resource Denise Pumain
    


    



    

    Latest news
    

    (There are no discussion topics yet in this forum)

    

    Recent activity
    

    Your last login was:
    Tuesday, 19 Aug 2003, 17:11
    

    

http://www.soc.surrey.ac.uk/nigel_gilbert.htm