Dear Applied Statistics Community,
Please join us Wednesday, November 19th, when Adam Glynn—Government
Department—will present his research, "Assessing the Empirical Evidence for
Mechanism Specific Causal Effects". Adam provided the following abstract:
Social scientists often cite the importance of mechanism specific causal
knowledge, both for its intrinsic scientific value and as a necessity for
informed policy. In this talk, I use counterfactual causal models to
re-assess
the empirical evidence for two oft cited examples from American and
comparative
politics: the voting habit effect that is not due to campaign attention and
the
effect of oil production on the likelihood of civil war onset that is due to
the weakening of state capacity. Utilizing decompositions of direct and
indirect effects, I discuss a number of identification strategies, and
demonstrate through sensitivity and bounding analysis that the evidence for
the
aforementioned examples is weaker than is typically understood.
The applied statistics workshop meets at 12 noon in room K-354, CGIS-Knafel
(1737 Cambridge St) with a light lunch. Presentations start at 1215 pm and
usually end around 130 pm. As always, all are welcome and please email me
with any questions
Cheers,
Justin Grimmer
Dead Applied Statistics Community,
Please join us this Wednesday, November 12th when Kosuke Imai will present
"Identification and Inference in Causal Mediation Analysis". Kosuke is
currently a professor in the Department of Politics at Princeton University
and an alum of the Harvard Government Department. He has provided the
following abstract for his talk:
Causal mediation analysis is routinely conducted by applied researchers in a
variety of disciplines
including communications, epidemiology, political science, psychology, and
sociology.
The goal of such an analysis is to investigate alternative causal mechanisms
by examining the
roles of intermediate variables that lie in the causal path between the
treatment and outcome
variables. In this paper, we first prove that under the assumption of
sequential ignorability,
the average causal mediation effects are nonparametrically identified. This
identification result
contrasts with previous studies which have concluded that the nonparametric
identification of
average causal mediation effects requires an additional assumption. Second,
we show that under
the same sequential ignorability assumption the average causal mediation
effects can be
identified in the linear structural equation model commonly used by applied
researchers. Some
practical implications of our identification result are also discussed.
Third, we consider a simple
nonparametric estimator of the average causal mediation effects and derive
its asymptotic
variance. Fourth, we offer sensitivity analyses in both parametric and
nonparametric settings
so that researchers can examine the robustness of their empirical findings
to the violation of the
sequential ignorability assumption. Finally, we analyze a randomized
experiment from political
psychology to illustrate the proposed methods.
A paper for the talk is available here:
http://imai.princeton.edu/research/files/mediation.pdf
The applied statistics workshop meets at 12 noon in room K-354, CGIS-Knafel
(1737 Cambridge St) with a light lunch. Presentations start at 1215 pm and
usually end around 130 pm. As always, all are welcome and please email me
with any questions
Cheers
Justin