I´m also interested in it

Hi Jasmine,

In my experience, I found that NetDraw is pretty limited in its usefulness. There are a bunch of reasons for this. This includes having limited control over many aspects of the graph (such as the scaling of edge widths, e.g. making weak edges disproportionately small). Perhaps most importantly I could only ever get it to create raster images rather than vectorised graphics. The latter are much better because if you put the network into a pdf it scales really nicely (you can zoom in). For example, if you zoom into the network on page 6 of my paper Farine & Milburn 2013 Social organisation of Thornbill-dominated mixed-species flocks using social network analysis in Behavioural Ecology and Sociobiology you can see how much better a vectorised network looks. 

I haven't got any experience with KUMU, but I know that there are two pretty good free options out there.

The first, which I use, is an open-source program called GEPHI. This creates really nice looking networks with lots of control over different aspects of it. It also allows you to calculate some basic node properties as well as communities. Its a bit tricky to import the data sometimes, and if you give it a go I'm happy to help with any hiccups (the online help files are pretty useful). Easiest way I found is to import via a csv (e.g. exported from Excel). Trick is that you have to replace all of the commas with semi-colons (;). I always include the node IDs as the first column and first row (thus the first value remains blank). Then you can easily import node attributes by having a column called 'ID' in the data table and the ids matching those in the network file. Once you're in Gephi its pretty smooth running and it can make really useful networks (see Figure 1 of Aplin et al 2013 Individual personalities predict social behaviour in wild networks of Great Tits in Ecology Letters, and the supplementary movie in Aplin et al 2015 Experimentally induced innovations lead to persistent culture via conformity in wild birds in Nature, both of which were created using Gephi).

The other option is to use the igraph package in R. This has a bit steeper learning curve, especially because the documentation for plotting networks is a bit limited. However, you can control just about every aspect of your network diagram, which is great! Also, if you are working in R, then you can do pretty much all of your analysis using igraph (or alternatively sna, another R package). Installation of these packages is super simple. If you are interested in going down this route, I have some code in Appendix 2 of my recent paper (Farine & Whitehead 2015 Constructing, Conducting and analysing animal social networks in Journal of Animal Ecology) that will get you underway and plotting your network in no time.

At the moment I use R for my network analysis and Gephi to plot them. However, I've got some projects for which I think R will be better suited for plotting (e.g. plotting a network over the top of a map).

Either way, I think if you take the plunge to move to a better solution you won't regret it. Good luck!

Damien

HI Damien,

Thanks for your input. YOur insights are much appreciated.  In a previous round of SNA software investigations i tried out gephi  before settling on ucinet/netdraw for a few minor display reasons.  I have not yet played with R but it sounds like i should as plotting a network over the top of a background image would be perfect for my project.

For note, in NetDraw you can 'save diagram as' / 'metafile' - this not only gives a vectorised graphic, but if you open it in an office program, say powerpoint, it can be converted to a microsoft office drawing object with all nodes, edges, arrow heads, text etc becoming editable objects and text boxes.  Infinitely editable and zoomable. Also, as it becomes a drawing object made of a group of shapes and text boxes without a 'background' it is easy to place an image or background behind such a network, again using powerpoint is easiest and later PDFable..

The challenges i am having in deciding between KUMU and Ucinet/netdraw at this stage are metrics based.  I wonder if anyone has any experience of using the metrics available in KUMU, in particular networks with directed ties.  

Jasmine

I would use Kumu for almost everything and if I need more complex algorithms for community detection I would switch to igraph in R.