The debate about the methodological problems surrounding influence effects in networks (see my previous post) continues on the SOCNET discussion list, mostly in reaction to Lyons’ critique of Christakis and Fowler’s well-known obesity studies. Noteworthy is a lengthy but thoughtful contribution by Tom Snijders, who argues that although Lyons’ points seem valid, no method is perfect when applied in practice, and we’ll have to make do with what we have and try to improve things in the meantime. I fully agree. See also here and here.