Based on the r “bipartite” package i would also recommend to try qgraph for additional colorful ways to represent networks. For example i really like the circle layout as displayed here (http://tinyurl.com/8b6nbgq). Greetings
I am wondering if it is possible(makes sense) to use 2-mode Bipartite network for continuous variables to identify possible profiles (clusters) of variables and subjects. for example this study http://www.skbhavnani.com/DIVA/papers/Bhavnani_et_al-JBI-2011.pdf, used Bipartite with Kamada-Kawai algorithms. . I tried to replicate their approach even using pre-classified datasets (continuous variables) and I couldn’t able to observe the patter I did expect; just it gave me random pattern. May be I am missing something!
I would be very grateful if you could kindly suggest me if it is possible to use Bipartite for that kind of research questions!
Best regards
Michael
Hi Michael, it’s hard to know what you might be missing based on your comment alone. I think stackoverflow.com or statsexchange.com are probably better places to get specific questions like this answered, but have a look at this posting guide, see if you can provide a reproducible example, and I’ll take a look.
I just wanted to drop a line to say thanks for posting this great article & tutorial! I’d been finding it hard to get started converting a bipartite graph to a social network (as a graph) and your post is truly appreciated!
… and then compute transition probabilities (the probability that a member of one group will stay a member of the same group or become a member of a new group…
Where is the calculation of the transition probability?
Hey there Hui, thanks for the comment. Right I don’t cover that here and I’ve removed the text you’re referencing. But you can find code to compute transition probabilities and explanations in our SNA labs: http://sna.stanford.edu/lab.php?l=5.
How would you approach a situation with n-mode model? Let’s say there are multiple ways people can connect (i.e. elementary school, junior high school, high school, college). In the graph, you might have 4 “root” nodes the represent each school. You would have smaller “people” nodes to represent students, and vertices to represent connections. A person could be associated with any number of schools and people.
Now you want to understand how students connect from elementary school to college. How do you structure the data and visualize it?
Well a lot depends on the question you really want to answer. If it’s just how people are connected, I’d say use a person-school edgelist to represent the data, loop over it to create a series of 1-mode networks (people-to-people), maybe join them, and visualize it with a simple network diagram. Maybe use different colors for edges corresponding to different schools.
That color coding is a clever method for the example above, but what about when you increase complexity? In my case, I have 4 or 5 unique ways students could be connected. That could be school, church, baseball team, etc.
I feel like color coding loses its value when you have large, nominal data sets. Is there some other method of showing how these groups connect? Maybe line thickness representing number of ways people connect? Or maybe a network graph isn’t the best solution.
Pingback: Working with Bipartite/Affiliation Network Data in R | Statistics with R | Scoop.it
Based on the r “bipartite” package i would also recommend to try qgraph for additional colorful ways to represent networks. For example i really like the circle layout as displayed here (http://tinyurl.com/8b6nbgq). Greetings
Pingback: Working with Bipartite/Affiliation Network Data in R | R | Scoop.it
I am wondering if it is possible(makes sense) to use 2-mode Bipartite network for continuous variables to identify possible profiles (clusters) of variables and subjects. for example this study http://www.skbhavnani.com/DIVA/papers/Bhavnani_et_al-JBI-2011.pdf, used Bipartite with Kamada-Kawai algorithms. . I tried to replicate their approach even using pre-classified datasets (continuous variables) and I couldn’t able to observe the patter I did expect; just it gave me random pattern. May be I am missing something!
I would be very grateful if you could kindly suggest me if it is possible to use Bipartite for that kind of research questions!
Best regards
Michael
Hi Michael, it’s hard to know what you might be missing based on your comment alone. I think stackoverflow.com or statsexchange.com are probably better places to get specific questions like this answered, but have a look at this posting guide, see if you can provide a reproducible example, and I’ll take a look.
Pingback: Working with Bipartite/Affiliation Network Data...
Pingback: Visualize & cluster one-mode projection of a bi-partite graph | Aakhayan
I just wanted to drop a line to say thanks for posting this great article & tutorial! I’d been finding it hard to get started converting a bipartite graph to a social network (as a graph) and your post is truly appreciated!
Yes, just a note of thanks, this was a really helpful tutorial.
… and then compute transition probabilities (the probability that a member of one group will stay a member of the same group or become a member of a new group…
Where is the calculation of the transition probability?
Hey there Hui, thanks for the comment. Right I don’t cover that here and I’ve removed the text you’re referencing. But you can find code to compute transition probabilities and explanations in our SNA labs: http://sna.stanford.edu/lab.php?l=5.
Hey Solomon,
How would you approach a situation with n-mode model? Let’s say there are multiple ways people can connect (i.e. elementary school, junior high school, high school, college). In the graph, you might have 4 “root” nodes the represent each school. You would have smaller “people” nodes to represent students, and vertices to represent connections. A person could be associated with any number of schools and people.
Now you want to understand how students connect from elementary school to college. How do you structure the data and visualize it?
Well a lot depends on the question you really want to answer. If it’s just how people are connected, I’d say use a person-school edgelist to represent the data, loop over it to create a series of 1-mode networks (people-to-people), maybe join them, and visualize it with a simple network diagram. Maybe use different colors for edges corresponding to different schools.
That color coding is a clever method for the example above, but what about when you increase complexity? In my case, I have 4 or 5 unique ways students could be connected. That could be school, church, baseball team, etc.
I feel like color coding loses its value when you have large, nominal data sets. Is there some other method of showing how these groups connect? Maybe line thickness representing number of ways people connect? Or maybe a network graph isn’t the best solution.
Pingback: Collapsing a bipartite co-occurrence network – Mubashir Qasim