That entirely depend on your raw data. R is great for manipulating data: have a look at Quick-R, http://www.statmethods.net/, for an introduction.

Best,
Tore

Tore

I am trying to use tnet for a behavioral dataset as a trial, I tried an undirected network .

sampledata <- rbind(
c(1,2,3),
c(1,3,2),
c(2,3,4),
c(2,4,5),
c(2,5,2),
c(6,5,1))

and used the weighted_richclub_w() function on the above. I find that for directed=FALSE and for rich="k" and reshuffle="weights", it gives the following error message:

"Error in random.m[i, j + 1] <- random.phi[which(random.phi[, "x"] == j), :
subscript out of bounds"

The same error appears for rich="k" and reshuffle="links" with directed=FALSE

With directed=NULL there is no error !

1. I don't understand why it gave an error with directed=FALSE even though the input was actually an undirected network.

2. For the case of directed weighted network, how is the generalization done for null models weight reshuffle and weights & links reshuffle of your PRL 101 paper ?

3. Does the combination rich="k", reshuffle="links" correspond to weight reshuffle, and rich="k", reshuffle="weights" correspond to weights & links reshuffle and rich="s", reshuffle="weights.local" correspond to directed reshuffle null model of your PRL 101 paper ?

Your comments on these will help me apply tnet to my data.

ciao,
jay

The standard answer is to use an edgelist as most R-packages can easily load this format. An edgelist is a list of the ties between nodes. For example:

1 2
3 1

would be the two ties from node 1 to node 2 and from node 3 to node 1. My package, tnet, uses a third column that differentiates the ties from one another:

1 2 4
3 1 2

so here, the first tie is twice as strong as the second tie.

To load a text file (.txt) with an edgelist, simply write in R:

net <- read.table(“filename_of_edgelist_file.txt”)

That said, the standard network analysis packages might not be appropriate for you as they generally are cross-sectional and cannot distinguish between the first tie and the last tie. From you description, you might want to keep this temporal aspect in your analysis. Also, you might want to consider keeping the two-mode structure (i.e., people and companies) instead of simply creating a network among people (see the post on projecting networks).

All the best,
Tore

I find your site extremely helpful. In my PhD I am working on a study of 20 entrepreneurial high tech firms in Iceland and building a framework of their international ventures. As the interviews with the firms have proceeded I have noticed aspects which have drawn my attention to social networks. Especially interesting was that many of the entrepreneurs mentioned the same individuals who were connectors in their initial international ventures. I am also observing their relations on the new social networks on the web and the networks which they talk about in the interviews and the networks on the web are quite different! This would be interesting to observe with SNAS tools. I have now a list of 130 individuals/links (all numbered from 1 to 130). Now I am struggling to set it up in an R-Framework. Any suggestions?

Best regards

Thor
University of Iceland

Glad you find this page useful.

Mark Newman sent me the two-mode network, which did contain the names of the authors. I have uploaded this file now (see the network description). Also, in the description of the Southern Women dataset is a link to the women’s first name.

All the two-mode network files list people ids first, and then the paper/event ids.

Hope this helps,

Tore

Also, this blog looks great!!

Regards,
Nav.