Hi Reut, sorry your email got stuck in the queue. If you are still having a problem, feel free to send me the data (sborgatti@...). You can send either the ucinet files
(don’t forget to include both the ##h and ##d files) or Excel.
?
steve
?
toggle quoted message
Show quoted text
From: [email protected] < [email protected]> On Behalf Of
Reut Liraz via groups.io
Sent: Thursday, May 19, 2022 06:49
To: [email protected]
Subject: [ucinet] E-I index calculation - Ego Networks. #egonetworks
?
?
Hello,?
I am trying to calculate ego-alter similarity via UCINET, however, although my net and attribute data are completely symmetrical on the Excell sheet? (both contain 192 rows), on the UCINET I receive different rows' sizes.
When I input the files into Netdraw it works smoothly, without any warning.
However, trying to calculate the E-I index, I receive an "Attribute vector must be the same size as matrix rows" notice.
Maybe someone can help solve this issue,
Thank you
Reut Liraz
|
Hello,?
I am trying to calculate ego-alter similarity via UCINET, however, although my net and attribute data are completely symmetrical on the Excell sheet? (both contain 192 rows), on the UCINET I receive different rows' sizes.
When I input the files into Netdraw it works smoothly, without any warning.
However, trying to calculate the E-I index, I receive an "Attribute vector must be the same size as matrix rows" notice.
Maybe someone can help solve this issue,
Thank you
Reut Liraz
|
Series of LINKS workshops on SNA this month
Hi all, just letting you know that the LINKS Center at the University of Kentucky is offering four workshops on SNA in June. These include both Windows UCINET-based and R-based workshops. The R-based workshops are designed to complement the new book by Borgatti, Everett, Johnson and Agneessens (2022). The workshops are online-only. Track 1. ?- June 6, 7, 8 Led by ?, ?and , this course provides a basic introduction to the theory and method of network research. ?Topics include theoretical perspectives, social capital, and the nuts and bolts of doing a network analysis using UCINET and Netdraw software (both Windows programs). Starting June 6, it meets Monday, Tuesday, and Wednesday from 9:30am-11:30am EDT and 12:30pm-2:30pm EDT (12 contact hours) and provides homework assignments and access to TAs. ?The course costs $1500 ($750 for students). Track 2. ?- June 13, 14, 15, 16 Taught by ,??this is a more technical and in-depth workshop focusing on the concepts and methods of SNA, particularly as they apply to specific research objectives. The mathematics behind the measures is explained, as well as how to use the measures in practice. UCINET for Windows software is used extensively. The course meets four times starting June 13: Monday, Tuesday, Wednesday, ?and Thursday 10:00am-12:00pm EDT and 12:30pm-2:30pm EDT (16 contact hours). A fifth day is led by the TAs. ?The course costs $1500 ($750 for students). Track 3. ?- June 9, 10, 13, 14, 16 Taught by ?and . Like Track 2, this is a more technical and in-depth workshop than the intro workshop, focusing on the concepts and methods of SNA, particularly as they apply to specific research objectives. This workshop uses the software package R, rather than UCINET. The mathematics behind the measures is explained, as well as how to use the measures in practice. The course meets five times across two weeks starting June 9 with each instructional day consisting of two sessions: 10:00am-12:15pm EDT and 12:45pm-3:00pm EDT (22.5 contact hours). Participants will receive homework,?which includes running analyses and interpreting results, and which they can perform in small groups of 2?or 3. These results are?then discussed at?the next meeting.? The course costs $2000 ($1000 for students). Track 4. ?- June 20, 21, 22, 27, 28 (pm only), 29 (pm only) Taught by ?and , this course covers ERGMs and SAOMs, two families of statistical models used to model the presence or absence of ties. While the course is introductory, prior familiarity with both statistics and network analysis is strongly advised. Starting after Track 2?and Track 3?finish, the course meets six?days (22.5?contact hours), across two weeks, starting June 20. The course aims to be interactive, using breakout sessions for the exercises and time between classes to consolidate knowledge. Participants will receive homework?involving?running further analyses and interpreting result. The homework can be?performed in small groups of 2?or 3. These results are?then discussed at?the next meeting. The course costs $2000 ($1000 for students). Registration , and will remain open until 48 hours prior to the start of each track or when the track reaches capacity. Registrants may cancel for a full refund up to 24 hours before the start of their track(s). ?Students receive a 50% discount. ?Additional (half price) discounts are available for members of the University of Kentucky community. ? Software It is useful to install the software before the workshop, so that you can get any difficulties ironed out. Please visit our . Contact The website for this workshop is . ?If you have any questions, please contact Scott Soltis <scott.m.soltis@...>. ?In addition, visit your specific track to get contact information for your instructors and TAs. ? ? ? ?
|
Drawing graphs without alter-to-alter ties
Hi Everyone,
I am trying to show some ego connections to different alters in NetDraw, is there a way to model only ego's ties? Thanks for your time :)
Ozzie
|
Re: Zeros in Matrices (QAP Correlations)
Thank you for continuing the analogy, Steve - I understand the logic (and appreciate the wit). Even though a zero could be there, it doesn't mean that it should be there.
Regards,
Christine
|
Re: Zeros in Matrices (QAP Correlations)
0 success, lol! In my opinion, you have to use missing values. The IV is number of games. That is legitimately zero for some pairs. The DV is a characteristic
of the games played. Imagine that the DV was how exciting the games were on a scale of 1 to 10. Since there were no games observed, it would be impossible to measure the DV.
Because your actual DV is a count, it looks as if a zero makes sense, since no tie-breakers occur when there are no games. But that is also a logical certainty
– an artifact – that should not be allowed to inflate your correlation. It’s like saying, if you want fewer cases of covid, do less testing. ?
?
steve
?
toggle quoted message
Show quoted text
From: [email protected] <[email protected]>
On Behalf Of CHRISTINE NEWTON via groups.io
Sent: Wednesday, May 25, 2022 15:47
To: [email protected]
Subject: Re: [ucinet] Zeros in Matrices (QAP Correlations)
?
?
Thanks kindly for your reply, Jeremy -?I have completed binomial tests using a subset of my data with a different research question, so I appreciate your suggestion and will explore its application to the research question that led to the
tennis club example/analogy.
In the meantime, I'm still at a loss regarding the best way to approach QAP correlation with matrices that have legitimately empty components (i.e., where zero is a meaningful quantity rather than a non-response or non-relationship - should I add zeros or leave
blank), and correlation with sparse matrices that contain many zeros. In my mind, I'm -? likely erroneously - envisioning that there may be a 'real' correlation coefficient that reflects the actual data points and an 'artificial' correlation coefficient that
reflects the presence of those extra zeros in the matrices. If leave those 0 cells blank, the significance level changes so perhaps there's a real and artificial significance level as well.? I would love to find some specific literature on this but have had
'zero' success so far.
Thanks again for your response -
Christine
|
Re: Zeros in Matrices (QAP Correlations)
Thanks kindly for your reply, Jeremy -?I have completed binomial tests using a subset of my data with a different research question, so I appreciate your suggestion and will explore its application to the research question that led to the tennis club example/analogy.
In the meantime, I'm still at a loss regarding the best way to approach QAP correlation with matrices that have legitimately empty components (i.e., where zero is a meaningful quantity rather than a non-response or non-relationship - should I add zeros or leave blank), and correlation with sparse matrices that contain many zeros. In my mind, I'm -? likely erroneously - envisioning that there may be a 'real' correlation coefficient that reflects the actual data points and an 'artificial' correlation coefficient that reflects the presence of those extra zeros in the matrices. If leave those 0 cells blank, the significance level changes so perhaps there's a real and artificial significance level as well.? I would love to find some specific literature on this but have had 'zero' success so far.
Thanks again for your response -
Christine
|
Re: Zeros in Matrices (QAP Correlations)
Hi Christine,
Your intuition is correct that adding zeroes will raise the correlation coefficient (and probably the significance) because all of the zeroes are correlated perfectly with each other.
That said, it's not clear why you would be using a QAP correlation for this. It seems like a bernoulli (or binomial) outcome, with the research question focusing on the proportion of games that go to tie-break sets. Why not fit a binomial model that reflects the data-generating process?
Best, Jeremy
On Tuesday, May 24, 2022, 08:28:29 PM GMT, CHRISTINE NEWTON <c.newton@...> wrote:
Greetings,
I have a question regarding zeros in matrices that will be correlated using QAP Correlation procedure.
I've created a simple example (attached) for illustration: Four players in a (very, very small!) tennis club. Nodes are player names. First variable is total number of games played. Second variable is the number of games played that involved a tie-breaker set. There are some players who have not played each other. I'd like to correlate the variables using QAP Correlation.
Here are the data for the matrices:
Alan played Charles 4 times in total, with 2 games going into a tie-breaker set.
Barb played Charles 2 times in total, with 0 games going into a tie-breaker set.
Charles played Don 6 times in total, with 1 game going into a tie-breaker set.
Don played Alan 3 times in total, with 1 game going into a tie-breaker set.
If I organize these data into columns in Excel and correlate the columns, the correlation coefficient is 0.48. However, simple vector correlation uses assumptions that don't hold when doing network analysis, so we use QAP correlation procedures that involve permutation of the matrices.
First attempt at QAP Correlation: If I organize these data into adjacency matrices and do NOT recode missing elements to zero (and add missing rows and force symmetry), the UCINET QAP Correlation is 0.4781 and the significance is 0.4215. The QAP Correlation coefficient is the same as the Excel correlation coefficient.
Second attempt at QAP Correlation: If I organize these data into adjacency matrices and I DO recode missing elements to zero (and add missing rows and force symmetry), the UCINET QAP Correlation is 0.7311 and the significance is 0.1634. Have the additional zeros in the previously empty matrix elements resulted in a stronger association between the two matrices?
The blank cells in the first QAP Correlation attempt aren't missing data - any zeros in the matrices are legitimate. But if I more zeros in the matrices, then the correlation coefficient increases and the p-value decreases...implying that the presence of those zeros are influencing the correlation coefficients upwards and the significance value downwards.
Thus, if I have a larger sparse network (e.g., 100 actors) with a lot of valid (empty or zero) cells, my adjacency matrices would be dominated by zeros in cells. Is the QAP correlation an appropriate procedure to use if I want to measure the strength of association of these sparse matrices? Won't the prevalence of zeros in larger matrices "skew" the correlation results even more than the little tennis club example?
Hoping that someone can provide clarification for me - I haven't been able to find resources about correlating sparse network matrices containing many legitimate zeros.
Regards,
Christine Newton
(Athabasca University, Canada)
|
Zeros in Matrices (QAP Correlations)
Greetings,
I have a question regarding zeros in matrices that will be correlated using QAP Correlation procedure.
I've created a simple example (attached) for illustration: Four players in a (very, very small!) tennis club. Nodes are player names. First variable is total number of games played. Second variable is the number of games played that involved a tie-breaker set. There are some players who have not played each other. I'd like to correlate the variables using QAP Correlation.
Here are the data for the matrices:
Alan played Charles 4 times in total, with 2 games going into a tie-breaker set.
Barb played Charles 2 times in total, with 0 games going into a tie-breaker set.
Charles played Don 6 times in total, with 1 game going into a tie-breaker set.
Don played Alan 3 times in total, with 1 game going into a tie-breaker set.
If I organize these data into columns in Excel and correlate the columns, the correlation coefficient is 0.48. However, simple vector correlation uses assumptions that don't hold when doing network analysis, so we use QAP correlation procedures that involve permutation of the matrices.
First attempt at QAP Correlation: If I organize these data into adjacency matrices and do NOT recode missing elements to zero (and add missing rows and force symmetry), the UCINET QAP Correlation is 0.4781 and the significance is 0.4215. The QAP Correlation coefficient is the same as the Excel correlation coefficient.
Second attempt at QAP Correlation: If I organize these data into adjacency matrices and I DO recode missing elements to zero (and add missing rows and force symmetry), the UCINET QAP Correlation is 0.7311 and the significance is 0.1634. Have the additional zeros in the previously empty matrix elements resulted in a stronger association between the two matrices?
The blank cells in the first QAP Correlation attempt aren't missing data - any zeros in the matrices are legitimate. But if I more zeros in the matrices, then the correlation coefficient increases and the p-value decreases...implying that the presence of those zeros are influencing the correlation coefficients upwards and the significance value downwards.
Thus, if I have a larger sparse network (e.g., 100 actors) with a lot of valid (empty or zero) cells, my adjacency matrices would be dominated by zeros in cells. Is the QAP correlation an appropriate procedure to use if I want to measure the strength of association of these sparse matrices? Won't the prevalence of zeros in larger matrices "skew" the correlation results even more than the little tennis club example?
Hoping that someone can provide clarification for me - I haven't been able to find resources about correlating sparse network matrices containing many legitimate zeros.
Regards,
Christine Newton
(Athabasca University, Canada)
|
Centralities As Directedness
Can UCINET use a symmetric matrix to create a matrix of centralities for all the nodes or just per row? If so, which procedure as choosing centrality now I just obtain per node. I am trying to study ways to create directedness in a 1-mode matrix that originated as a 2-mode matrix. It would be nice if could be done with a variety of centralities to study those effects too. Thank you.
|
Re: About two-mode analysis on xUCINET and UCINET
Check out the difference here. Generally I go with NMDS but try both and see which ordination method gives results that are better fits. It's amazing how much variance can be captured in so few dimensions.
|
It also runs fine on Win 11 running under Parallels on a Mac.
|
Hi Christian, good to hear from you. For those of you teaching with UCINET, the following exercises and tutorials might be helpful.
?
·???????
. These are tersely worded but take you through typical analyses. They are appropriate for
both current and older versions of ucinet.
·???????
. These are more verbose and detailed documents that are currently under construction. Some
of these rely on datasets found in ucinet 6.747 (the latest version) or later. (They are actually copies of old dataset with nicer new names. For example, the old ‘zackar’ and ‘zachattr’ files are now ‘karatenet’ and ‘karateattr’. Similarly, ‘krack-high-tec’
and ‘high-tec-attributes’ are now ‘kracknet’ and ‘krackattr’.);
?
steve
?
toggle quoted message
Show quoted text
From: [email protected] < [email protected]> On Behalf Of
Christian Steglich via groups.io
Sent: Tuesday, May 17, 2022 18:00
To: [email protected]
Subject: Re: [ucinet] windows 11
?
?
Hi Steve,
I can also confirm that it works well. I am currently teaching with UCINET as companion software on my Windows 11 Surface.
I regret the Windows 11 update, but that's a different issue.
Best, Christian
?
On 17-May-22 15:53, Andrea Salvini wrote:
Dear Steve,
I am using Windows 11 on daily basis and I confirm that Ucinet runs smoothly on this system.
Best
Andrea Salvini
?
Inviato da per Windows
?
?
Hi, a potential customer wants to know if UCINET runs on Windows 11. I assume it does, but don’t have a Windows 11 machine to confirm. Any of you running UCINET on Windows 11?
?
thanks
?
steve
?
Stephen P. Borgatti
Gatton Endowed Chair of Management
Gatton College of Business and Economics
University of Kentucky
sborgatti@...
?
--
Interuniversity Centre for Social Science Theory & Methodology
Department of Sociology, Grote Rozenstraat 31, NL-9712 TG GRONINGEN
|
Hi Steve,
I can also confirm that it works well. I am currently teaching
with UCINET as companion software on my Windows 11 Surface.
I regret the Windows 11 update, but that's a different issue.
Best, Christian
On 17-May-22 15:53, Andrea Salvini
wrote:
Dear Steve,
I am using Windows 11 on daily basis and I
confirm that Ucinet runs smoothly on this system.
Best
Andrea Salvini
?
Inviato da per Windows
?
?
Hi, a potential customer
wants to know if UCINET runs on Windows 11. I assume it
does, but don’t have a Windows 11 machine to confirm. Any of
you running UCINET on Windows 11?
?
thanks
?
steve
?
Stephen P. Borgatti
Gatton Endowed Chair of
Management
Gatton College of
Business and Economics
University of Kentucky
sborgatti@...
?
--
Interuniversity Centre for Social Science Theory &
Methodology
Department of Sociology, Grote Rozenstraat 31, NL-9712 TG
GRONINGEN
|
Hi, ? I am using Windows 11 and Ucinet runs. ? Best Regards, ? ? ? Lara Bertola ?
toggle quoted message
Show quoted text
From: Steve BorgattiSent: 17 May 2022 16:53 To: [email protected]Subject: Re: [ucinet] windows 11 ? Thanks Paulo and Andrea. ? ? Dear Steve, I am using Windows 11 on daily basis and I confirm that Ucinet runs smoothly on this system. Best Andrea Salvini ? Inviato da per Windows ? ? Hi, a potential customer wants to know if UCINET runs on Windows 11. I assume it does, but don’t have a Windows 11 machine to confirm. Any of you running UCINET on Windows 11? ? thanks ? steve ? Stephen P. Borgatti Gatton Endowed Chair of Management Gatton College of Business and Economics University of Kentucky sborgatti@... ?
|
Thanks Paulo and Andrea. ?
toggle quoted message
Show quoted text
From: [email protected] < [email protected]> On Behalf Of Andrea Salvini Sent: Tuesday, May 17, 2022 09:53 To: [email protected]Subject: Re: [ucinet] windows 11 ? Dear Steve, I am using Windows 11 on daily basis and I confirm that Ucinet runs smoothly on this system. Best Andrea Salvini ? Inviato da per Windows ? ? Hi, a potential customer wants to know if UCINET runs on Windows 11. I assume it does, but don’t have a Windows 11 machine to confirm. Any of you running UCINET on Windows 11? ? thanks ? steve ? Stephen P. Borgatti Gatton Endowed Chair of Management Gatton College of Business and Economics University of Kentucky sborgatti@... ?
|
Dear Steve, I am using Windows 11 on daily basis and I confirm that Ucinet runs smoothly on this system. Best Andrea Salvini ? Inviato da per Windows ? ? Hi, a potential customer wants to know if UCINET runs on Windows 11. I assume it does, but don’t have a Windows 11 machine to confirm. Any of you running UCINET on Windows 11? ? thanks ? steve ? Stephen P. Borgatti Gatton Endowed Chair of Management Gatton College of Business and Economics University of Kentucky sborgatti@... ?
|
Hi, until now ucinet worked without problems in w11.? (-:
toggle quoted message
Show quoted text
Hi, a potential customer wants to know if UCINET runs on Windows 11. I assume it does, but don’t have a Windows 11 machine to confirm. Any of you running UCINET on Windows 11?
?
thanks
?
steve
?
Stephen P. Borgatti
Gatton Endowed Chair of Management
Gatton College of Business and Economics
University of Kentucky
sborgatti@...
?
|
Hi, a potential customer wants to know if UCINET runs on Windows 11. I assume it does, but don’t have a Windows 11 machine to confirm. Any of you running UCINET on Windows 11?
?
thanks
?
steve
?
Stephen P. Borgatti
Gatton Endowed Chair of Management
Gatton College of Business and Economics
University of Kentucky
sborgatti@...
?
|
Re: About two-mode analysis on xUCINET and UCINET
Hi Martin, Thanks!
May I ask two more questions about the UCINET functions?
1. What’s the difference between metric and non-metric MDS? Can I assume that the former is for valued networks and the latter for binary data? 2. For node-level hypothesis testing, if I want to perform logistic regression, is it possible to do so??
Thanks!?
-Chih-Hui
Chih-Hui LAI (赖至慧), PhD Associate Research Fellow Research Center for Humanities and Social Sciences?(RCHSS) Academia Sinica Taipei, Taiwan, 115 Email: imchlai@... ? ?? Tel: 886-02-27898130 Mobile: 0966-192-712 Web:
toggle quoted message
Show quoted text
Hi I vaguely recall an error in the Davis data that was corrected and this could cause the change. Also we do update and improve routines and correct bugs and these too could make a difference.? Martin? ? Sent from??for Windows ? ? Hi Martin, Thanks! That’s really helpful. I tried the dual-projection method, following the steps mentioned in the document. There are indeed 5 core events now. But the women-by-women partition generated 10 core members, rather than 8 as described in the book (Figure 13.5, Analyzing Social Networks). I followed the settings (including Minres as algorithm) when doing the one-mode continuous core-periphery. The transformed one-mode data and the output are as attached. Not sure what caused this discrepancy? ? Chih-Hui LAI ( 赖至慧), PhD Associate Research Fellow Research Center for Humanities and Social Sciences?(RCHSS) Academia Sinica Taipei, Taiwan, 115 Email: imchlai@...?? ?? Tel: 886-02-27898130 Mobile: 0966-192-712 Web: ??
? Hi Dual core periphery in UCINET uses the categorical method whereas xUCINET uses the dual projection which is a continuous method. The later and usually better method is the dual projection. This can be done in UCINET but requires a few extra steps but gets the same results see? for details how to do this inUCINET. Martin
On 4 May 2022, at 03:58, Chih-Hui Lai <chlai@...> wrote:
?Dear all,
I’ve used UCINET to run Davis_SouthernWomen data, using 2-mode categorical core-periphery. As the attached output shows, the core events are 7, 8, and 9.?
But these results seem different from the ones if using xUCINET (via xDualCorePeriphery function), which showed that the core events are 5,6, 7,8 and 9.
xDualCorePeriphery(Davis_SouthernWomen$Attendance)
[[1]]
??EVELYN ????LAURA ??THERESA ???BRENDA CHARLOTTE ??FRANCES ??ELEANOR ????PEARL ?????RUTH ????VERNE ????MYRNA KATHERINE?
???????1 ????????1 ????????1 ????????1 ????????0 ????????0 ????????0 ????????0 ????????0 ????????0 ????????0 ????????0?
??SYLVIA ?????NORA ????HELEN ??DOROTHY ???OLIVIA ????FLORA?
???????1 ????????1 ????????0 ????????0 ????????0 ????????0?
[[2]]
E01 E02 E03 E04 E05 E06 E07 E08 E09 E10 E11 E12 E13 E14?
?0 ??0 ??0 ??0 ??1 ??1 ??1 ??1 ??1 ??0 ??0 ??0 ??0 ??0?
Does anyone know the reason about the difference?
Thanks!
-Chih-Hui
Chih-Hui LAI (赖至慧), PhD
Associate Research Fellow
Research Center for Humanities and Social Sciences (RCHSS)
Academia Sinica
Taipei, Taiwan, 115
Email: imchlai@...?????
Tel: 886-02-27898130
Mobile: 0966-192-712
Web:?
<ucinetlog4.txt>
?
|
Paulo Matui