Keyboard Shortcuts
Likes
Search
Question on the statistical analysis of SNA data
¿ªÔÆÌåÓýHi all ¨C I have a question about statistically analyzing SNA results.? The situation is this: there is a group of 11 investors who each fall into one two types, ¡°A¡± or ¡°B¡± and who can invest jointly with any other investor (the "invests_with" event is non-directional).? I¡¯d like to know if either group tends to invest more with their own type or with the other type, that is: "A:A" joint investments vs. "A:B" joint investments vs. "B:B" joint investments. I¡¯m OK at applying SNA but I¡¯m not a statistician, and this situation is more complicated because there are different numbers in each group type (there are 2 ¡°A¡± investors and 9 ¡°B¡± investors).? Also, each individual investor has jointly invested a different number of times; the table below shows how many times each investor has invested with each other investor.? Given this brief background, is it possible to determine if the ¡°A¡± and ¡°B¡± investors tend to invest more their own group or with the other group?? If it¡¯s possible, how do I do it? Thanks for any help you can give!
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Dear Paul, I would strongly suggest you to invest a few hours on Stephen Borgatti, Martin Everett, Jeffrey Johnson, "Analyzing Social Networks" 2013, which is a gentle introduction to SNA with Ucinet examples. It certainly cover the topics you are?interested in: related research design, data preparation and analysis. I think no statistical background is required in order to fully understand most of the chapters.? Following standard approaches - including these authors' one - I would suggest to represent investors by name (or anonymous ID)? in the adjacency matrix and to distinguish A and B as attributes for these investors in a separate attribute vector. Everything is gently covered by the book. Hope this helps. Good work ! Teresio
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
There are a number of ways to do it. A good one is qap correlation. You have a valued network:
A B B B A B B B B B B A 10 0 0 1 5 3 7 3 0 1 B 10 0 0 9 0 11 8 10 0 0 B 0 0 0 0 0 0 0 0 0 0 B 0 0 0 0 0 1 1 0 0 0 A 1 9 0 0 0 6 4 5 1 1 B 5 0 0 0 0 3 4 1 0 0 B 3 11 0 1 6 3 9 14 0 0 B 7 8 0 1 4 4 9 8 0 0 B 3 10 0 0 5 1 14 8 0 1 B 0 0 0 0 1 0 0 0 0 0 B 1 0 0 0 1 0 0 0 1 0 And a categorical attribute: Type A 1 B 2 B 2 B 2 A 1 B 2 B 2 B 2 B 2 B 2 B 2 (only 2 As, which is not much to work with). You can convert this to a ¡°is the same type as¡± matrix using Data|Affiliations): A B B B A B B B B B B A 1 0 0 0 1 0 0 0 0 0 0 B 0 1 1 1 0 1 1 1 1 1 1 B 0 1 1 1 0 1 1 1 1 1 1 B 0 1 1 1 0 1 1 1 1 1 1 A 1 0 0 0 1 0 0 0 0 0 0 B 0 1 1 1 0 1 1 1 1 1 1 B 0 1 1 1 0 1 1 1 1 1 1 B 0 1 1 1 0 1 1 1 1 1 1 B 0 1 1 1 0 1 1 1 1 1 1 B 0 1 1 1 0 1 1 1 1 1 1 B 0 1 1 1 0 1 1 1 1 1 1 And then run Qap correlation: This yields something like this: QAP results for paulnet * paulatt-sameType (5000 permutations) 1 2 3 4 5 6 7 8 9 Obs Value Significa Average Std Dev Minimum Maximum Prop >= O Prop <= O N Obs --------- --------- --------- --------- --------- --------- --------- --------- --------- Pearson Correlation -0.1443 0.2935 -0.0003 0.2176 -0.3377 0.4357 0.7259 0.2935 5000.0000 The results are non-significant. Steve. From: ucinet@... <ucinet@...> Sent: Sunday, November 18, 2018 17:45 To: ucinet@... Cc: Dr. Paul Beckman <pbeckman@...> Subject: [UCINET] Question on the statistical analysis of SNA data Hi all ¨C I have a question about statistically analyzing SNA results. The situation is this: there is a group of 11 investors who each fall into one two types, ¡°A¡± or ¡°B¡± and who can invest jointly with any other investor (the "invests_with" event is non-directional). I¡¯d like to know if either group tends to invest more with their own type or with the other type, that is: "A:A" joint investments vs. "A:B" joint investments vs. "B:B" joint investments. I¡¯m OK at applying SNA but I¡¯m not a statistician, and this situation is more complicated because there are different numbers in each group type (there are 2 ¡°A¡± investors and 9 ¡°B¡± investors). Also, each individual investor has jointly invested a different number of times; the table below shows how many times each investor has invested with each other investor. Given this brief background, is it possible to determine if the ¡°A¡± and ¡°B¡± investors tend to invest more their own group or with the other group? If it¡¯s possible, how do I do it? Thanks for any help you can give! A B B B A B B B B B B A - 10 0 0 1 5 3 7 3 0 1 B - 0 0 9 0 11 8 10 0 0 B - 0 0 0 0 0 0 0 0 B - 0 0 1 1 0 0 0 A - 0 6 4 5 1 1 B - 3 4 1 0 0 B - 9 14 0 0 B - 8 0 0 B - 0 1 B - 0 B - |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Hi Steve,
Looking at this example you provided to Paul, I am curious about a small adjustment. Your solution answers the question "do like people invest with each-other more than with the other group." But you could have a situation where As invest with As, but Bs do not necessarily invest with Bs (or vice versa). So if you wanted to test that you could make "is A" and "is B" affiliation matrices and presumably run a QAP regression, where the "is A" and "is B" matrices are explanatory variables. My question is, would you need to make any kind of adjustment when interpreting the significance of the results? It seems like this is rather analogous to an ANOVA with post hoc two-way tests in which you would do something like a bonferroni adjustment. Is it correct to assume that no such adjustment is needed with the QAP regression as this is taken into account by the regression itself? Best wishes Jesse On Wed, Nov 21, 2018 at 3:29 PM 'Steve Borgatti' steve.borgatti@... [ucinet] <ucinet@...> wrote:
-- *Jesse Sayles, PhD* *saylesunchartedwaters.com <>* *ORISE Postdoctoral Fellow* *Appointed with the U.S. Environmental Protection Agency, Office of Research and DevelopmentNational Health and Environmental Effects Research Laboratory, Atlantic Ecology Division* |