ORA lost 300 edges during import

[Semenoffalex]

Hi!

I've stumbled upon the following problem:

I have a *.csv file with 1951 edges. 2 of them are self-loops. Imported it to ORA 2.3.6 on Windows 7 through "Simple table" wizard option. It sees only 1650 links and 2 self-loops.

NodeXL showed the following:

Unique Edges 1445 
Edges With Duplicates 506 
Total Edges 1951  

Self-Loops 2 

I've attached the file.

Any Ideas?

[Miray Kas]

For instance, there are multiple entries of

 

temkinmir;Alexgizn

 

It might be combining them to a single edge with higher weight. Please try checking the link weights.

 

Thanks,

Miray

 

 

2012/1/27 Semenoffalex <semeno...@gmail.com>

[Semenoffalex]

Yep, but that doesn't explain huge discrepancies in in-degree
centralities between ORA and NodeXL =(

On 27 янв, 21:05, Miray Kas <miray...@gmail.com> wrote:
> For instance, there are multiple entries of
>
> temkinmir;Alexgizn
>
> It might be combining them to a single edge with higher weight. Please try
> checking the link weights.
>
> Thanks,
> Miray
>

> 2012/1/27 Semenoffalex <semenoffa...@gmail.com>

[Kathleen Carley]

You need to check your preferences - e.g., did you binarize, use
weights, symmetrize ....

[Semenov Alexander]

It's exactly what I did and the results were quite strange. There are only 2 self-loops, but analysis results with and without them differ too much.

Here is in-degree centrality for weighted, asymmetric, network with 2 self-loops:

navalny 76

olly_jolly 56

temkinmir 27

xenia_sobchak 26

burmatoff 24

Alexgizn 16

varlamov 15

tvrain 12

VRSoloviev 11

adagamov 10

And here is in-degree centrality for weighted, asymmetric network without 2 self-loops:

navalny 76

Alexgizn 64

temkinmir 63

olly_jolly 57

Tractorcoff 37

angeina_may 30

xenia_sobchak 26

burmatoff 24

YanaBogdanova 19

varlamov 15

I don't understand, how removing 2 self-loops can change the results so much and why they still doesn't match with the ones from NodeXL?

2012/1/30 Kathleen Carley <kathlee...@cs.cmu.edu>

[geoffrey...@gmail.com]

ORA uses the weighted centrality formulas, and does not assume link symmetry.

Try binarizing and symmetrizing your ORA network (right-click on the network and hit transform) and then recalculating your measures.


Sent on the Sprint® Now Network from my BlackBerry®

[Semenov Alexander]

After binarizing the results finally matched.

Thanks, Geoff.

2012/1/30 <geoffrey...@gmail.com>