The Social Science Statistics Blog points to the latest in a series of papers examining the relationship between social ties and various health issues. While I think this research is important in that it raises the profile of networks; however, the consistent failure to support research with actual network analysis significantly weakens the results.
The implicit assertion in this research is that logistic regression on social ties is a form of network analysis, from which network effects can be measure. In my view, this is wholly false, and is equivalent to the notion that network visualization alone is network analysis. While this may seem pedantic, consider the consequences for the research referenced above.
The logistic regression estimates, for example, the relationship on height and relationships in dyad pairs. This explains nothing about what is driving individuals to organize this way; leaving several critical unanswered questions. Are tall people actually clustering around one another, or day they simply have dyadic relationships with other tall people? What type of mixing occurs within the structure, or are these clusters exclusive? Where are the critical nodes, and how do they distinguish themselves, and what effect does height have on that (in this example)? Only a detailed examination of the network structure will illuminate these questions.
I, therefore, pose this question to the researchers in this burgeoning field: how can you claim to measure network effects, without network analysis?
Automatically Generated Related posts:
Popular Posts