the compliers and defiers are, and so who the local average treatment effect
applies to is a nebulous subset of the sample.
Finally, the variables you use as your treatment or your outcome variable
might only be proxies for what you are truly interested in. For instance,
though you may be interested in looking at whether economic shocks push
people to commit suicide, data on suicides may not be available (or suicides
may be significantly under-reported), and so
you might have to resort to
using mortality rates instead.
Dostları ilə paylaş: