So I’ve been working with the National Longitudinal Study of Adolescent Health (Add Health) for a while but I’ve only recently began looking at the raw friendship nomination data. I’m hoping that someone can give me some practical advice.
My first question this: would you recommend using the network or igraph package?
I’m working in R, and I want to create some measures of centrality. I wasn’t planning on doing ERG models or anything else complicated at the moment, just simple stuff. If you want to recommend a different programming environment I’m happy to hear you make your case.
Permutations 
In R I prefer igraph, but if you are only interested in looking at the Ad-Health data statnet—which subsumes the network package—is probably the better choice since it comes with simulated versions of Ad-Health network data (see faux.mesa.high and faux.magnoila.high in the ergm package).
Click to access ergm.pdf
since friendship nomination is directed, you might want to use igraph as it can do alpha centrality. as far as i know (and i’m probably wrong), statnet doesn’t have any centrality indices that are optimized for directed graphs. although i’m not an R person, i’ve generally been pretty impressed by igraph.
I am currently using the statnet package and it works fairly well. Although, I encountered problems when comparing R centrality results to other software results. Ideas on this issue would be appreciated.
i’m transitioning more and more to R, and likely to iGraph (we’re having Csardi here to give a talk & workshop on it in a couple of weeks). A few thoughts from my limited use of the packages in R i know (though i honestly still do most of my analyses in SAS packages Jim Moody wrote).
i’ve used network/SNA a bit (which as Drew said are now subsumed in statnet) and i believe it has directed closeness and betweenness. i work with diseases so much that i often don’t think about others like eigenvector, etc. (though i bet one or two others are in SNA as well).
Personally, i find that statnet handles data more similarly to the ways i’m used to than iGraph (which is why i’m looking forward to our workshop), but that may just be inertia. Alternately, while i can get graphs out of network/SNA quickly, i feel like i can more easily tweak them in iGraph. My $0.02.
Phillip, the differences across packages are generally in how the values are being normalized. You can typically tweak those in the options.
Thanks for the feedback everyone! I should bleg more often.
If anyone has any more thoughts on statnet vs. igraph do share.