Close
Close

Not a magic wand

Picking an application for conducting network analysis

In this article, the fourth in a series on network analysis, Arianna reviews three software applications that she has tried for her research. She will explain the reasons why she opted to use ORA.

Written by Arianna Sacco on

In the previous entries to this series, I have written about what network analysis is, and choices that you need to make concerning the study of networks. I have also shown the first steps of network analysis, namely compiling the database and the matrix.

The next step is importing the matrix in the software program for network analysis. Therefore, I thought it would be useful to compare three programs that you can get for free. These are Gephi, VISONE, and ORA. I have tried all three before finally opting to stick with ORA for my own research.

Before continuing, I would like to say that this article is very much based on my personal experience. The software applications I mention here are just three of all the available ones: they are suggested for network analysis based on archaeological finds.

There are, of course, other programs out there (e.g. Pajek), some of which you need to pay for (e.g. UCINET). In other fields, such as social sciences or research based on written documents, other programs would probably be more suitable than the ones I’ve used.

Of course, what you will end up using is your own choice and depends on your goals and on your data: there is no wrong or right choice, it’s just about what is more suited to your particular situation.

Exploring the software

Both Gephi and VISONE use side drop-down menus and wizards to show the different functions that allow you to analyse and visualize the networks. This allows the programs to display at the same time both the graph and the functions you can perform on the data.

In contrast, ORA makes a larger use of drop-down menus, and the network is shown in a different window than the one where you conduct the analysis. However, in ORA you can easily visualize, and modify if necessary, the matrix you have imported. Even though all the programs allow you to easily modify features, I have found ORA more comfortable to use, but it’s just my personal preference.

As far as the measures – i.e. the algorithms that examine the role of an element in the network form a mathematical point of view – are concerned, Gephi offers fewer algorithms than the other programs. ORA offers more algorithms than VISONE, but VISONE gives more options as far as fine-tuning them is concerned.

Especially when it comes to detecting the so-called modularity, namely when you want to divide the elements in groups or “families” based on of many features they have in common, I find that VISONE is probably better, because it gives more options when it comes to the algorithms available and to the options for fine-tuning the analysis. In my case, VISONE gave clearer results when I wanted to explore how the sites could be grouped based on the similarity of their objects, or to detect if determined types of objects were more characteristic of particular sites.

However, as I will explain in a future article, I decided to follow a different procedure to analyse the overall similarity of the objects examined from the different sites, thus I did not need the modularity anymore. All the programs described here offer an overview of how the elements of the network (in my case: the sites) score for the measures, so that you can compare and contrast them.

However, VISONE gives more options to manipulate and perform advanced operations with the data. For example, you can scale the values of the links, or normalize them based on a specific attribute, or add an offset, or a threshold. Both VISONE and Gephi also offer many options to filter out elements of the network based on specific parameters, with many options to fine-tune these parameters. While using ORA, I’ve had the impression that these options were fewer, even though ORA gives options for statistical analysis, such as correspondence analysis, which are not provided for by the other programs.

In certain instances, to deepen the analysis, or just for completeness’ sake, it is useful to add further attributes, namely features that characterize the elements of the network. For example, in my case a feature could be the colour of the beads, or the shape of the head and body of the scarabs, or the geographic coordinates and the dating of the sites.

I found that adding – and modifying – attributes is easier in Gephi and in ORA than in VISONE. This also makes it more convenient to create networks based on geographic distances, because the data needed can quickly be added and re-worked. I find that this, and the GIS (Geographical Information System) plugin that ORA provides, make ORA useful for archaeological networks, which often have a geographical connotation.

As far as visualizing the graphs is concerned, I think that Gephi is the best option, because it gives many possibilities when it comes to colouring and sizing the nodes (the elements examined), the links (the features they share), and their labels: this allows you to really show the subtleties concerning the strength of the links and the role of the nodes. Gephi also gives many options when it comes to the layout of the networks, namely the structure, or shape, of the network, with many types of layouts you can choose from, and many layout features that you can fine-tune.

Even though ORA offers many options for layout too, you cannot do much fine-tuning. In general, I think that larger sets of data are better visualized in Gephi, though this didn’t concern my specific case, because my dataset wasn’t tremendously large.

Opting for ORA

In the end, I opted for ORA because of four main advantages, which were important to my research and which the other programs did not provide. Warning: ORA has a paid version and a trial (free) version that you need to re-download every six months.

Firstly, ORA provides the option to easily binarize a network. In my previous article, I have explained that my two-mode networks were binary. Despite that, the original matrix for my two mode-networks was not binary, but reported the number of contexts, whenever available, because that was information that I did not want to lose. But for the analysis that I wanted to conduct, a binary network was more applicable. ORA gave me the possibility of both keeping a weighted matrix and analysing a binary network, because I could just binarize the matrix after importing it in the program.

Secondly, ORA gives the option to easily transforming two-mode networks into one-mode networks (I talked about this in the aforementioned article), with just a click, and keeps both the two-mode and the one-mode networks in the same file, allowing you to analyse all of them at the same time.

Thirdly, ORA allows you to work on multiple networks at the same time and to easily perform operations involving more networks. For example, I could create a new network where both beads and stone vessels were present, by simply adding these two networks, or I could easily intercept the elements that these networks share (for example a site where particular types of beads and of stone vessels are found together).

Finally, ORA gives the possibility, through a GIS plugin, to visualize the network on a map. With the other programs, all these things are not possible, or are less easy to do.

Closing remarks

To sum up, then, Gephi provides a better visualization of the network and allows you to get a good visual impression of it. VISONE allows for a deeper analysis and is especially good when it comes to distinguish specific groups in the network and “playing around” with the data. ORA also allows for a deeper analysis and is especially useful when you are working with multiple types of networks and when your study has a geographical element to it.

As I mentioned earlier in this article, this is not to say that one program is somehow better than the other. Every program has stronger and weaker features, what you choose ultimately depends on which features you need more in your research.

Top