Re: Analysis of egonet data: Can I use forfiles to join multiple output datasets?
#egonetworks
Hi Malcolm, I’m able to select multiple files in data|join|columns, like this:
?
?
Maybe a keyboard issue?
?
steve
?
toggle quoted message
Show quoted text
From: [email protected] < [email protected]> On Behalf Of
Malcolm Alexander via groups.io
Sent: Tuesday, February 15, 2022 21:47
To: [email protected]
Subject: [ucinet] Analysis of egonet data: Can I use forfiles to join multiple output datasets? #egonetworks
?
?
Hi Steve (and group)
I often get asked how to deal with multiple egonet datasets. Forfiles in CLI will run a function like cohesion over multiple files as long as the names are consistent (i.e. egonet1, egonet2....).
This produces more multiple files/datasets; egonet1-coh, egonet2-coh... (as explained in Perry, Pescosolido and Borgatti)
I want to aggregate them into one dataset (i.e. Joincols) in order to run Univariate Stats across the whole array.
The join columns procedure in the main menu (-> Data ->Join) requires one to select each output file one at a time which is tedious when one has lots of cases. Similarly the CLI joincols requires each file name to be entered.
I have experimented with the forfiles procedure with joincols but cannot make it do this aggregation (concatenation).
Any suggestions?
Cheers,
Malcolm.
PS: Does the forfiles procedure still accept the ? for a single character wild card? It didn't seem to work for me, I had to use the *.
|
Re: Line length in NetDraw
#help
Tom,
Only in 4 dimensions.? In two dimensions draw a circle around an one point that passes through all the other four points. It’s clear that these other four points cannot be equally distant from one another..? ??
?
toggle quoted message
Show quoted text
From: [email protected] < [email protected]> On Behalf Of
Tom Gallagher
Sent: Wednesday, February 16, 2022 8:54 AM
To: [email protected]
Subject: [ucinet] Line length in NetDraw #help
Importance: Low
?
Hello UCINET community,
I have been working in NetDraw in an attempt to visualize rather basic networks (such as the one in the picture below).
I was wondering if there is a way to make all of the lines in this network equal length for visualization purposes. The closest thing I have found is Layout-->Circle, but that still does not make all of the lines equal in length.
Let me know if there is any way I can do this.
Thank you!
|
Line length in NetDraw
#help
Hello UCINET community, I have been working in NetDraw in an attempt to visualize rather basic networks (such as the one in the picture below). I was wondering if there is a way to make all of the lines in this network equal length for visualization purposes. The closest thing I have found is Layout-->Circle, but that still does not make all of the lines equal in length. Let me know if there is any way I can do this. Thank you! 
|
Analysis of egonet data: Can I use forfiles to join multiple output datasets?
#egonetworks
Hi Steve (and group)
I often get asked how to deal with multiple egonet datasets. Forfiles in CLI will run a function like cohesion over multiple files as long as the names are consistent (i.e. egonet1, egonet2....).
This produces more multiple files/datasets; egonet1-coh, egonet2-coh... (as explained in Perry, Pescosolido and Borgatti)
I want to aggregate them into one dataset (i.e. Joincols) in order to run Univariate Stats across the whole array.
The join columns procedure in the main menu (-> Data ->Join) requires one to select each output file one at a time which is tedious when one has lots of cases. Similarly the CLI joincols requires each file name to be entered.
I have experimented with the forfiles procedure with joincols but cannot make it do this aggregation (concatenation).
Any suggestions?
Cheers,
Malcolm.
PS: Does the forfiles procedure still accept the ? for a single character wild card? It didn't seem to work for me, I had to use the *.
|
Line size scaling UCINET
#help
Hello UCINET colleagues,
I have been working on visualizing a network in Netdraw, and am working on the finding the ideal scale for the size of lines based on their tie strength.
I have been somewhat confused on how the scaling works, and wanted to see if anyone could explain. When changing some of the min/max values, some of the line sizes appear to change while others do not, depending on what values I input. For example, the thinnest lines appear to stay exactly the same size when changing from a min of 1 and max of 15 to a min of 5 and max of 15. If anyone can explain how the scaling works, this would be hugely helpful.
Thank you!
|
Re: Entropy in Components
OK. It worked. Thanks professor Borgatti
El mié., 9 de febrero de 2022 3:03 p.?m., Steve Borgatti < sborgatti@...> escribió:
Jorge, is should be the natural log – (i.e., ln).
?
steve
?
?
?
Good morning UCINET community:?
I?m working with components (see the?report below)
How is estimated Entropy ? I tried log function with base 2 and 10 but the results are different
Can someone help me?
--
|
Jorge Enrique Mejía Quiroga
Profesor Asociado
|
|
Facultad Ingeniería y Ciencias Básicas
Dirección de Investigación y Transferencia Conocimiento
|
|
|
|
|
|
|
|
|
Acreditación Institucional de Alta Calidad. Resolución del MEN n.° 256 de 2019, vigente por 4 a?os. Vigilada Mineducación.
|
|
|
Alerta legal:?Este correo electrónico no representa la opinión
o el consentimiento expreso de la Universidad Central. La información aquí contenida es confidencial y puede tener la condición de mensaje privilegiado, el cual no puede ser usado ni divulgado a personas distintas de su destinatario. Se prohíbe su retención,
uso, grabación, aprovechamiento o divulgación con cualquier propósito. Si por error recibe este mensaje, devuélvalo a su remitente y elimínelo de su correo.
|
?
|
|
|
|
Re: Entropy in Components
Jorge, is should be the natural log – (i.e., ln).
?
steve
?
toggle quoted message
Show quoted text
From: [email protected] <[email protected]>
On Behalf Of Jorge Mejía via groups.io
Sent: Wednesday, February 9, 2022 12:25
To: [email protected]
Subject: [ucinet] Entropy in Components
?
?
Good morning UCINET community:?
I?m working with components (see the?report below)
How is estimated Entropy ? I tried log function with base 2 and 10 but the results are different
Can someone help me?
--
|
Jorge Enrique Mejía Quiroga
Profesor Asociado
|
|
Facultad Ingeniería y Ciencias Básicas
Dirección de Investigación y Transferencia Conocimiento
|
|
|
|
|
|
|
|
|
Acreditación Institucional de Alta Calidad. Resolución del MEN n.° 256 de 2019, vigente por 4 a?os. Vigilada Mineducación.
|
|
|
Alerta legal:?Este correo electrónico no representa la opinión
o el consentimiento expreso de la Universidad Central. La información aquí contenida es confidencial y puede tener la condición de mensaje privilegiado, el cual no puede ser usado ni divulgado a personas distintas de su destinatario. Se prohíbe su retención,
uso, grabación, aprovechamiento o divulgación con cualquier propósito. Si por error recibe este mensaje, devuélvalo a su remitente y elimínelo de su correo.
|
?
|
|
|
|
Re: Calculating Q modularity index from RCT numbers
Nicely done! I found it annoying that your steps were not quite automatable because divconst would only accept a numeric constant. So I fixed that (in ucinet
6.740, not available yet). So now you can do this
?
->run mal.txt lltab
?
with the attached text file mal.txt (assumed to be in the working directory).
?
Of course, the built-in modularity function is more efficient, but as you say, it does nothing for understanding the metric.
?
steve
?
?
?
toggle quoted message
Show quoted text
From: [email protected] <[email protected]>
On Behalf Of Malcolm Alexander via groups.io
Sent: Tuesday, February 8, 2022 22:23
To: [email protected]
Subject: Re: [ucinet] Calculating Q modularity index from RCT numbers
?
?
Hi Steve,
Thanks for pointing me to the equation. The first formulation seems to vary but the reduction to Trace of observed - sum of e squared is in several places. The version I used (attached) can be explained as the uniform distribution of joint probabilities summed
by column (incoming 'stubs') by all cells in a row (outgoing) for each row of the probability matrix.
The Trace - Sum is much easier to demonstrate in the CLI.
->E = divconst(lltab,190)
->TrObs = sum(getdiagonal(E))
Diagonal included in calculations. To exclude, add diag:no to command.
Create E squared (matrix not element multiplication, Product)
Diagonal included in calculations. To exclude, add diag:no to command.
->Q = sub(TrObs,EsqudSum)
1 rows, 1 columns, 1 levels.
|
Good morning UCINET community:?
I?m working with components (see the?report below)  How is estimated Entropy ? I tried log function with base 2 and 10 but the results are different
Can someone help me?
-- Jorge Enrique Mejía Quiroga
Profesor Asociado | Facultad Ingeniería y Ciencias Básicas
Dirección de Investigación y Transferencia Conocimiento
| | | | | | Acreditación Institucional de Alta Calidad. Resolución del MEN n.° 256 de 2019, vigente por 4 a?os. Vigilada Mineducación. |
| | Alerta legal:?Este correo electrónico no representa la opinión o el consentimiento expreso de la Universidad Central. La información aquí contenida es confidencial y puede tener la condición de mensaje privilegiado, el cual no puede ser usado ni divulgado a personas distintas de su destinatario. Se prohíbe su retención, uso, grabación, aprovechamiento o divulgación con cualquier propósito. Si por error recibe este mensaje, devuélvalo a su remitente y elimínelo de su correo. |
| |
|
Re: Calculating Q modularity index from RCT numbers
Hi Steve, Thanks for pointing me to the equation. The first formulation seems to vary but the reduction to Trace of observed - sum of e squared is in several places. The version I used (attached) can be explained as the uniform distribution of joint probabilities summed by column (incoming 'stubs') by all cells in a row (outgoing) for each row of the probability matrix. The Trace - Sum is much easier to demonstrate in the CLI.
Create E
->E = divconst(lltab,190)
->TrObs = sum(getdiagonal(E))
Diagonal included in calculations. To exclude, add diag:no to command.
?
Create E squared (matrix not element multiplication, Product)
->Esqud = prod(E,E)
Sum Esqud
->EsqudSum = sum(Esqud)
Diagonal included in calculations. To exclude, add diag:no to command.
?
Calculate Q?
?
->Q = sub(TrObs,EsqudSum)
->dsp Q
?
? ? ? ? ? ? ? ? ?1?
? ? ? ? ? ? ? ?sum?
? ? ? ? ? ? ? ----?
? ? ?1 temp#1 0.096?
?
1 rows, 1 columns, 1 levels.
?
|
Re: Question on edgelist to matrix
What you have shown are the eigenvalues not the eigenvector scores. Scroll down further to see the centrality values.
toggle quoted message
Show quoted text
On 7 Feb 2022, at 19:18, Steve Borgatti <sborgatti@...> wrote:
?
Aakanksha, if you like, send me your data (sborgatti@...) and I will see if there is anything wrong. If you send the ucinet dataset, don’t forget to include both the
##h and ##d files. Or send the edgelist.
?
steve
?
?
?
Hi Steve,?
I did as you suggested but unfortunately, the eigen vector centrality measures are completely out of the range, for example, for node number 1 it is something like 20.something like this:
? ? ? ? ? ?Eigenv Pct Va Cum Pc?
? ? ? ? ? ? ?alue riance? ? ? t?
? ? ? ? ? ?------ ------ ------?
? ? ? ? ?1 20.049? 5.067? 5.067?
? ? ? ? ?2? 7.832? 1.979? 7.046?
? ? ? ? ?3? 5.775? 1.459? 8.505?
? ? ? ? ?4? 3.447? 0.871? 9.376?
? ? ? ? ?5? 3.162? 0.799 10.175?
? ? ? ? ?6? ? ? 3? 0.758 10.934?
? ? ? ? ?7? ? ? 3? 0.758 11.692?
? ? ? ? ?8? ? ? 3? 0.758 12.450?
? ? ? ? ?9? 2.767? 0.699 13.149?
? ? ? ? 10? 2.709? 0.685 13.834?
? ? ? ? 11? 2.709? 0.685 14.518?
? ? ? ? 12? 2.709? 0.685 15.203?
? ? ? ? 13? 2.558? 0.646 15.849?
? ? ? ? 14? 2.499? 0.632 16.481?
Seems completely odd, I also want to know how can i calculate network level and cluster level measures?
Thanks, Aakanksha
|
Re: Question on edgelist to matrix
Aakanksha, if you like, send me your data (sborgatti@...) and I will see if there is anything wrong. If you send the ucinet dataset, don’t forget to include both the
##h and ##d files. Or send the edgelist.
?
steve
?
toggle quoted message
Show quoted text
From: [email protected] < [email protected]> On Behalf Of
neha21gaur via groups.io
Sent: Monday, February 7, 2022 04:46
To: [email protected]
Subject: Re: [ucinet] Question on edgelist to matrix
?
?
Hi Steve,?
I did as you suggested but unfortunately, the eigen vector centrality measures are completely out of the range, for example, for node number 1 it is something like 20.something like this:
? ? ? ? ? ?Eigenv Pct Va Cum Pc?
? ? ? ? ? ? ?alue riance? ? ? t?
? ? ? ? ? ?------ ------ ------?
? ? ? ? ?1 20.049? 5.067? 5.067?
? ? ? ? ?2? 7.832? 1.979? 7.046?
? ? ? ? ?3? 5.775? 1.459? 8.505?
? ? ? ? ?4? 3.447? 0.871? 9.376?
? ? ? ? ?5? 3.162? 0.799 10.175?
? ? ? ? ?6? ? ? 3? 0.758 10.934?
? ? ? ? ?7? ? ? 3? 0.758 11.692?
? ? ? ? ?8? ? ? 3? 0.758 12.450?
? ? ? ? ?9? 2.767? 0.699 13.149?
? ? ? ? 10? 2.709? 0.685 13.834?
? ? ? ? 11? 2.709? 0.685 14.518?
? ? ? ? 12? 2.709? 0.685 15.203?
? ? ? ? 13? 2.558? 0.646 15.849?
? ? ? ? 14? 2.499? 0.632 16.481?
Seems completely odd, I also want to know how can i calculate network level and cluster level measures?
Thanks, Aakanksha
|
Structural holes measures
? I have my structural hole measures from Ucinet and would like to know what particular measurements to explore mynetwork structure I should use.
Can you create a variable from these measures or should I take all five components Degre Effsize Effic Const Hiera into consideration?
Are there any empirical articles you can recommend about this in order to develpo my hyphotesis? ? Lara Bertola ?
|
Re: Question on edgelist to matrix
Hi Steve,? I did as you suggested but unfortunately, the eigen vector centrality measures are completely out of the range, for example, for node number 1 it is something like 20.something like this:
? ? ? 1? ? ? 2? ? ? 3?
? ? ? ? ? ?Eigenv Pct Va Cum Pc?
? ? ? ? ? ? ?alue riance? ? ? t?
? ? ? ? ? ?------ ------ ------?
? ? ? ? ?1 20.049? 5.067? 5.067?
? ? ? ? ?2? 7.832? 1.979? 7.046?
? ? ? ? ?3? 5.775? 1.459? 8.505?
? ? ? ? ?4? 3.447? 0.871? 9.376?
? ? ? ? ?5? 3.162? 0.799 10.175?
? ? ? ? ?6? ? ? 3? 0.758 10.934?
? ? ? ? ?7? ? ? 3? 0.758 11.692?
? ? ? ? ?8? ? ? 3? 0.758 12.450?
? ? ? ? ?9? 2.767? 0.699 13.149?
? ? ? ? 10? 2.709? 0.685 13.834?
? ? ? ? 11? 2.709? 0.685 14.518?
? ? ? ? 12? 2.709? 0.685 15.203?
? ? ? ? 13? 2.558? 0.646 15.849?
? ? ? ? 14? 2.499? 0.632 16.481?
Seems completely odd, I also want to know how can i calculate network level and cluster level measures?
Thanks, Aakanksha
|
Re: Question on edgelist to matrix
I assume that by ‘numbered’ you mean that the node names are numerical ids. As far as ucinet is concerned, these are just names. So, you can add the missing nodes anywhere. Where they end up in the matrix is
determined by the ‘sort alphabetically’ checkbox:
?
?
If that is checked off, then no matter what order things were input, the matrix will be sorted from smaller id to larger id.
?
Just give it a try and see what if it does what you want it to.
?
steve
?
toggle quoted message
Show quoted text
From: [email protected] < [email protected]> On Behalf Of
Jeffrey Broadbent via groups.io
Sent: Friday, February 4, 2022 16:56
To: [email protected]
Subject: Re: [ucinet] Question on edgelist to matrix
?
?
If the edge list is numbered, do you have to add in the missing nodes at their proper numbered place, or can you add them at the bottom as you did with John?
Jeffrey Broadbent
Professor Emeritus,?
Fellow, Institute on the Environment
University of Minnesota
Email:
broad001@...
“The world is much more interesting than any one discipline.”? – Edward Tufte
?
?
On Thu, Feb 3, 2022 at 3:20 PM Steve Borgatti <sborgatti@...> wrote:
One way to do it is to add the missing nodes to the original edgelist as a tie of node to itself. For example, john is missing from this edgelist:
?
Bill Terry
Terry Bob
Tina Bill
?
So I add John to the bottom:
?
Bill Terry
Terry Bob
Tina Bill
John John
?
(actually, you don’t even need the second John – this would work tooJ
?
Bill Terry
Terry Bob
Tina Bill
John
?
?
?
?
?
?
?
If one converts an edgelist to a matrix, the rows and columns (nodes/actors) with all zero cells are eliminated from the produced matrix. ?what is the easiest?way to repopulate
those absent?rows and columns to make up the whole original?matrix? ?so far I have been doing it by hand.? Is there an automated way in UCINet?
Jeffrey Broadbent
Professor Emeritus,?
Fellow, Institute on the Environment
University of Minnesota
Email:
broad001@...
“The world is much more interesting than any one discipline.”? – Edward Tufte
|
Re: Question on edgelist to matrix
If the edge list is numbered, do you have to add in the missing nodes at their proper numbered place, or can you add them at the bottom as you did with John?
Jeffrey Broadbent
Professor Emeritus,? Department of Sociology Fellow, Institute on the Environment University of Minnesota Email: broad001@...“The world is much more interesting than any one discipline.”? – Edward Tufte
On Thu, Feb 3, 2022 at 3:20 PM Steve Borgatti < sborgatti@...> wrote:
One way to do it is to add the missing nodes to the original edgelist as a tie of node to itself. For example, john is missing from this edgelist:
?
Bill Terry
Terry Bob
Tina Bill
?
So I add John to the bottom:
?
Bill Terry
Terry Bob
Tina Bill
John John
?
(actually, you don’t even need the second John – this would work tooJ
?
Bill Terry
Terry Bob
Tina Bill
John
?
?
?
?
?
?
?
If one converts an edgelist to a matrix, the rows and columns (nodes/actors) with all zero cells are eliminated from the produced matrix. ?what is the easiest?way to repopulate those absent?rows and columns to make up the whole original?matrix?
?so far I have been doing it by hand.? Is there an automated way in UCINet?
Jeffrey Broadbent
Professor Emeritus,?
Fellow, Institute on the Environment
University of Minnesota
Email: broad001@...
“The world is much more interesting than any one discipline.”? – Edward Tufte
|
Re: Calculating Q modularity index from RCT numbers
Hi Malcolm, you can calculate Q several ways in ucinet. One way is the Tools|Cluster|Moca. Another way is in the CLI. For example, suppose I want to use the Zachary karate club data. There are two relations and three attributes. I can type: ? ->dsp modularity(zackar zachattr) and get this: ? Matrix: Q modularity ?????????????????? 1?????? 2?????? 3 ?????????????Support Strengt??? Club ???????????????????????????h???????? ?????????????------- ------- ------- ?????1 ZACHE? 0.3439 -0.0864? 0.3582 ?????2 ZACHC? 0.3810 -0.0786? 0.3914 ? 2 rows, 3 columns, 2 levels. ? To calculate Q, UCINET uses equation 4 here: ? steve ?
toggle quoted message
Show quoted text
From: [email protected] <[email protected]> On Behalf Of Malcolm Alexander
Sent: Friday, February 4, 2022 00:55
To: [email protected]
Subject: [ucinet] Calculating Q modularity index from RCT numbers ? I am using the Relation Contingency tables (Tools ->Testing Hypotheses ->Mixed Dyadic Nodal ->Categorical ->Relational Contingency Table) to see whether network partitions, such as those produced by the NCD, are statistically significant.
I am wondering if we can use the frequency of ties count that an RCT provides to calculate the Q modularity score for a particular partition. When I try to interpret Newman's formula for Q modularity I just get the E minus I index (reverse sign).
Given the importance of Q measures for several subgroup detection procedures I want to explain to my classes where it comes from.
Thanks for advice on this. As always, it is much appreciated.
|
Re: Question on edgelist to matrix
Hi, paste your data into the dl editor (4th button from left on the toolbar). Click into the “type” column and choose edit it delete it. change the data format to edgelist1, the select undirected. Also click ‘Ignore first row’. In other words, make it look like this: ? 
? When you press save, the data will be converted into a ucinet dataset which you can use to perform analyses like centrality. ? steve ?
toggle quoted message
Show quoted text
From: [email protected] <[email protected]> On Behalf Of neha21gaur via groups.io
Sent: Friday, February 4, 2022 05:01
To: [email protected]
Subject: Re: [ucinet] Question on edgelist to matrix ? Hi. how do i convert edge list into a matrix. something like this into a matrix ? does ucinet takes this form as a valid input to compute centrality? or it is mandatory to convert into matrix form? Source | Target | Type? | Weight | 1 | 299 | Undirected | 1 | 2 | 87 | Undirected | 1 | 3 | 423 | Undirected | 1 | 4 | 424 | Undirected | 1 | 5 | 425 | Undirected | 1 | 6 | 426 | Undirected | 1 | 7 | 427 | Undirected | 1 |
|
hi all i have a ties/edges table that looks like this, where 1, 2, 3...are the nodes. how can i convert them into a matrix form? i need to compute centrality. does ucinet allows to compute the metrics based on the format below or do i need to convert them into matrix form. If yes how do i do it? Kindly guide
| Source |
Target |
Type? |
Weight |
| 1 |
299 |
Undirected |
1 |
| 2 |
87 |
Undirected |
1 |
| 3 |
423 |
Undirected |
1 |
| 4 |
424 |
Undirected |
1 |
| 5 |
425 |
Undirected |
1 |
| 6 |
426 |
Undirected |
1 |
| 7 |
427 |
Undirected |
1 |
|
Re: Question on edgelist to matrix
Hi. how do i convert edge list into a matrix. something like this into a matrix ? does ucinet takes this form as a valid input to compute centrality? or it is mandatory to convert into matrix form?
| Source |
Target |
Type? |
Weight |
| 1 |
299 |
Undirected |
1 |
| 2 |
87 |
Undirected |
1 |
| 3 |
423 |
Undirected |
1 |
| 4 |
424 |
Undirected |
1 |
| 5 |
425 |
Undirected |
1 |
| 6 |
426 |
Undirected |
1 |
| 7 |
427 |
Undirected |
1 |
|
Malcolm Alexander
