In our first class we engaged with the most basic questions about how to adjust a treatment effect for potential confounders. We ran into the curse of dimensionality, wondered about whether adjustment based on functional form alone is useful, and produced a propensity score (although post-stratifying on it alone did not help solve the problems arising from the facts of the design and data). Here is the commented version of classwork 1: Classwork 1 with comments And here is the Sweave source code.