Contributions: Abstract
Causal inference is most easily understood using potential outcomes, which
include all post treatment quantities. The use of potential outcomes to
define causal effects is particularly important in more complex settings,
i.e., observational studies or randomized experiments with complications
such as noncompliance. Here we deal with the issue of estimating the
casual effect of a treatment on Quality of Life (QOL) in a randomized
experiment where some of the patients die before their QOL could be
assessed at the appropriate time. Because both QOL and death are
post-randomization quantities, they both should be treated as potential
outcomes. This perspective on QOL was first proposed by Rubin (1998), and
leads to the use of Principal Stratification (Frangakis and Rubin, 2002),
where the causal effect on QOL is well defined only for the stratum of
subjects who would live under either treatment assignment. Some results
will be presented from Zhang and Rubin (2004) and from current work on
involving job training programs. There are relationships to classical
"instrumental variables" models in economics for treating noncompliance
(Angrist et al., 1996), but IV's underlying assumptions are implausible in
the QOL setting, which inherently is more challenging.
|