Stanislav Markus writes:
Dave, Gary, Tao,
We are asked to do both matching and regression. Should we allocate our
energy equally between the two, say 5 pages each, or should we focus
more on regression, and do matching just to compare results?
Both matching and regression are equally important.
Also, we've never talked in-depth about matching
in non-experimental
settings.
I tried to! Recall the time we spent of the treatment effect of "being talked
to on the phone". Although it was true that being assigned to the treatment of
"being on the call list" was done randomly (more or less), the treatment
assignment of actually talking to someone was *not* random. The potential
voters who were talked to were very different from ones that were not talked
to. That is why we had to match.
What I wonder about is correct verbiage and
conceptual
clarity: is it meaningful e.g. to talk about incumbency as "treatment",
Yes. If you want to talk about the causal effect of incumbency than,
conceptually at least, it must be a treatment.
or about "an assignment mechanism" since
nobody really assigns
treatments here..?
As in the case of the "getting talked to" treatment, nature (or, rather, some
force other than experiment) does the assignment.
If this is not clear to you, I urge you to take a look at the Holland and Rubin
readings that I suggested.
Dave
Thank you,
Stan
--
David Kane
Lecturer In Government
617-563-0122
dkane(a)latte.harvard.edu
Please avoid sending me Word or PowerPoint attachments.
See
http://www.fsf.org/philosophy/no-word-attachments.html