Print

Print


*****  To join INSNA, visit http://www.insna.org  *****

Hi Ales and all-

Thank you for your help and advice.  Yes there would be unnecessary
redundancy in the scenario I described.  In fact, assigning nodal values
while reporting an edgelist would assign nodal values not twice, but every
time a node was involved in an edge (from once to many many times).
However, if the values are simply overwritten that redundancy would not
really be a problem.

Regardless, there is no need to introduce the nodal attributes at that stage
if it is easy to introduce partitions keyed to the node labels (not the
pajek assigned number) and if the vector need not be in the same order as
the nodes as introduced from the edgelist.

Is this true?  Can I introduce a vector of length N, ordered by label (and
not ordered by the pajek assigned number)?    I ask because the numbering
and ordering in your example seem to suggest otherwise.

I am trying to figure out the steps necessary to enter edges, edge weights
and nodal attributes-- here is an attempt:

Scenario 1:  (assuming one can introduce a partition vector ordered by
label)

Generate weighted edgelist; save as csv; import via Txt2Payek.
Generate node list with attributes; recode attributes into numeric values;
copy and paste attribute vector into notepad (for partition); below that
paste the vector of node labels; and then import that as a partition?

Scenario 2:  (cannot introduce vector ordered by label)   then this becomes
the copy/paste  scenario that I was hoping to avoid.

I can see creating a .paj file by hand.  But not for dozens of networks with
thousands of nodes.  So I am still hoping for a better solution.

thanks again,

ted







On 1/12/06, Ales Ziberna <[log in to unmask]> wrote:
>
> Hello!
>
> There is a very good reason why Pajek does not allow you to specify nodal
> attributes like this. That reason is that in your case you could specify
> two
> different values of an attribute for the same node.
>
> You should combine your nodal attributes in a vector, which can be also
> used
> as partition. I believe the vector should have numerical values. Any
> vector
> of the same length as the number of nodes can be used as nodal attributes
> in
> Pajek. If you want that nodes of different type would be ploted with
> different symbols, you can specify that too. For example, if you have:
> Sender  recipient  edge_weight  sender_type  recip_type
> 1       2         7          A           A
> 1      3          3          A           B
>
> You could put that in Pajek format (in *.paj file) as:
>
> *Network Example
> *Vertices 3
> 1 "Name 1" box
> 2 "Name 2" box
> 2 "Name 3" ellipse
> *Edges
> 1 2 7
> 1 3 3
>
> *Partition AB partition
> *Vertices 3
> 1
> 1
> 2
>
> *Vector AB vector
> *Vertices 3
> 1
> 1
> 2
>
>
> Putting those nodal attributes in a vector should not be a problem!
>
> Best,
> Ales Ziberna
>
> -----Original Message-----
> From: Social Networks Discussion Forum [mailto:[log in to unmask]] On
> Behalf Of Ted Welser
> Sent: Thursday, January 12, 2006 9:08 PM
> To: [log in to unmask]
> Subject: Re: edgelists into pajek, with edge weights and node attributes
>
> *****  To join INSNA, visit http://www.insna.org  *****
>
> Vlado-
>
> Thank you for your quick reply.  I am afraid that I do not see the bearing
> of "How to 6" on importing data.  Perhaps you meant "How to 5"?
>
> From "How to 5" I see that Txt2Pajek allows introduction of edge
> attributes-- including weight.  The other two inputs also refer to line
> information.  Is there a way use these input fields to specify node
> attributes (like interior color) for the nodes in Column 1 and Column
> 2?  I
> experimented a bit and the fourth and fifth fields generated edge data,
> not
> data for the verticies.
>
> If not, it seems like the TXT2Pajek tool would be much more powerful if it
> had a structure that allowed the user to specify whether a column refered
> to
> the tie or either vertex 1 or vertex 2 associated with that tie.
>
> Anyways-- I am hoping there is an easy short term solution to importing
> actor attributes as nodal information.
>
> thanks again,
>
> ted
>
> _____________________________________________________________________
> SOCNET is a service of INSNA, the professional association for social
> network researchers (http://www.insna.org). To unsubscribe, send an email
> message to [log in to unmask] containing the line UNSUBSCRIBE SOCNET
> in
> the body of the message.
>
>

_____________________________________________________________________
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.