Previous Entry Share Next Entry
05:25 pm, 21 Sep 03

graphing, social networks

The SNA Package
This is a fully documented collection of R routines for social network analysis; utilities included range from hierarchical Bayesian modeling of informant accuracy to logistic network regression (with QAP and CUG tests). Quite a few low-level utilities for plotting and transforming networks are available as well, along with many of the usual centrality and distance measures.

I keep trying to learn R, but I think I am missing some crucial intuition for making everything fit together. (Here’s a hint: they use the period in function names in the way they use underscores in other languages, as a spacer and *not* as a “member-of” operator.)

I want to graph the data I’ve been collecting on LiveJournal. I have unix timestamps and the time it took to load the LJ front page at that time. The hurdles are:
  1. I don’t have data for every time point, so I cannot simply plot load times and I must use the unix times for the x axis.
  2. I need to convert the unix times to real times upon display, but I need them as unix times (ie, plain integers) for graphing/comparison.
  3. The data is really bumpy and needs smoothing, and I don’t even know how to begin to approach that.