ABSTRACT
SAT Scores
for Sale? Pseudoexperimental
Assessment of Commercial Test Preparation via Optimal Full
Matching
Ben Hansen, Department of Statistics,
University of Pennsylvania
Stratification is an old, flexible,
efficient, and
conceptually plain statistical technique. If there is sufficient
pre-treatment data to discern which study subjects are comparable to
one another, and if every treated subject is sufficiently like some
control to justify comparison to it, then with the right
stratification one can rightly estimate treatment effects simply by
averaging and differencing outcomes.
The right stratification need not take a simple form, in which case
algorithms commonly in use won't find it. However, there is always
a
so-called full matching that is as good as any other possible
stratification. Full matchings enjoy various practical and
theoretical advantages over other matchings. Since no existing
statistical software creates full matchings, I am developing S
routines that do so easily; I shall illustrate their use by analyzing an
observational study of effects of commercial test preparation on
SAT scores.
Return to
2001-2002
Schedule of Presentations | Return to Home Page
|