Still Going

Apologies for the hiatus, I’ve spent the last few weeks immersed in data and at the beach, with little time for blogging either way.  This means, of course, that it’s time for another data update!

Specifically, I want to talk about one variable and one method that are turning out to be very interesting.  The method is a new test that I had never run before in UCINET, the social network analysis program.  This test is called an interactive ‘convergence of iterated correlations’ or CONCOR.  It looks at the actors (the people we interviewed and those identified in interviews) in the network, and divides them into groups based on structural similarities.  The test keeps dividing and dividing (hence the ‘iterated’) until the most similar actors are separated into different groups.  Fortunately, with UCINET, I can push a few buttons and see this happen in front of my eyes in about 2 seconds, instead of repeating statistical analyses over and over again by hand to get the same results.  Unfortunately, this means that when my project advisor Dr. Aday asks me what exactly are the statistical and mathematical formulas being used, my best answer is “…computer magic?”.

This test was very exciting not only because I figured out how to run a new useful procedure in a relatively complicated program, but also because the results of the test then highlighted gender as a key variable.  When the actors were divided into most similar groups based on their structural characteristics (who mentioned whom, how many times someone was mentioned or mentioned others, etc. etc.) the groups were almost completely divided by gender.

While I still need to dig into this finding to see the depth and robustness of the connection, using a new test to find some support for part of the original hypothesis, which I might not have found otherwise, feels like research at its best.