On DAGs, Instruments and Social Networks
Apropos the lively discussion we have had here on graphs and IV models, Felix Elwert writes:
“Dear Judea, here’s an IV paper using DAGs that we recently published. We use DAGs to evaluate genes as instrumental variables for causal peer effects in social networks. The substantive question is whether obesity is contagious among friends. We evaluated both single-IV and IV-set candidates. We found DAGs especially helpful for evaluating variations within a large class of qualitative data-generating processes by playing through a lots of increasingly realistic variations of our main DGPs (Table 1). Being able to discuss identification purely in terms of qualitative causal statements (e.g., “fat genes may affect latent causes of friendship formation”) was very helpful. One of our goals was to check far one can relax the DGP before identification would break down. After permitting a host of potential hurdles (i.e., eliminating exclusions left and right by drawing lots of additional arrows), we concluded that genes alone won’t realistically work, but that time-varying gene expression might give valid IVs for peer effects. Under linearity, we found suggestive evidence for the transmission of obesity in social networks.”
All the best,
Felix
I am trying to reconcile the work that Felix is describing with Imbens’s statement
that “in social science applications a graph with many excluded links may not be an attractive way of modeling dependence structures.” [Imbens; page 377]’
I do not know what Imbens meant to say about the unique features of social science applications;
perhaps that in such applications there are better ways of modeling dependence structures, or
that in such applications a graph with just a few excluded links may be a more attractive way of modeling dependence structures, or
that some social scientists prefer to avoid the modeling of dependence structures because they still do not know how to do it.
Whatever the meaning was, it does not seem to capture the work that most social scientists do, which requires the modeling of dependence structures.
Judea
Comment by judea pearl — November 17, 2014 @ 8:41 pm
There are talks that Huawei and LG will be working on the design of the upcoming Google Nexus 5 (2015) and it is expected to appear similar to the Nexus 6.
Comment by Hermes iPhone 6 + Case — May 28, 2015 @ 12:28 am