Advice for the rank beginner part 4 So when last we spoke, you had a picture in NetDraw. Maybe a mess, maybe you can already see something interesting. What do I mean by interesting? Well, is it a densely connected ball or are there lots of little groups sitting around. Hm.

Components. These are groups of nodes connected to each other but not to the rest of the graph. If you were looking at co-authorship within an organization, and you didn't have any collaboration across departments, each of these might represent a department. They might represent research topics or areas in either co-author or co-citation networks. Find these in NetDraw using Analysis > components > and then mark them by color or by shape. My co-authorship network looks festive:
If you want to see the components one at a time look over on the right hand side of netdraw... see a dropdown box with ID, change that to components, then see how you have the list of components? then below that a > 0 ! then below that letters a i s c ^D R. The s will step through the components. You can then probably even turn the labels back on when you get reasonable clusters. And you can use the spring embedding layout to re-arrange things again.

Ok, now lets look for nodes who are really well connected. On netdraw you can hit one button and calculate a bunch of different centrality measures. Analysis > centrality measures... of these, Degree is the most straightforward. A node's degree centrality is the number of connections it has. You might have some components that are complete - all of the connections that could exist do. So all of the nodes in that will have the same degree centrality. With me? Right. So then, well, this might just be the co-authors on the same paper. Well, Sitkis gave you line weights, too, so now you can see if there are two or more author1 - author2 links. You can make the lines heavier or put numbers next to them. Or you might have a star type thing. Or you might just have a mess with a few of high centrality.

I know a little bit (but not that much) more than I'm saying, but all I promised was advice for the rank beginner.

If you want to look at co-authors only, you can export your RefWorks database into a tab delimited file. Import into excel. Delete off everything but the authors. You'll need quotes around each author name since there's a space. You can use the concatenate command so you end up with:
"author fi mi" "author2"...
spaces between NOT tabs. Copy into your handy text editor so no weirdness. So then you can use the same header as you got from sitkis, but you need to know how many nodes. If you took your whole refworks database, you can add this up this pretty quickly using lookup by author.... Maybe there's a better way. In any case this would not be good in high energy physics - for obvious reasons.

Coming up: the conclusion... where to go for more info.  ¶ 8:35 PM| |cites (technorati) |