ABSTRACT
In the previous chapters, with the exception of factor analytic types of procedures, we have generally considered procedures that produce ‘end results.’ In other words,
models are estimated and tested and inferences are made from those results. In comparison, propensity score analysis (PSA) is a precursor to estimating outcomes, the results from which can be applied to any number of procedures for examining causal outcomes. For cooking enthusiasts, this could be considered akin to marinating. The marinade is something applied prior to the cooking procedure, a tool used to help produce a better outcome. Likewise with PSA. The theme of this chapter is how to conduct PSA. The procedures from many of the previous chapters can then be applied to answer the substantive efficacy questions that PSA is often created for. In this chapter, we will, however, focus on getting the data prepared-through propensity score analysis-for that substantive analysis. In the propensity score process, there are a number of analytic decisions that must be made when conducting PSA, and several new foundational concepts that come with this. Like other procedures covered in the text, propensity score matching is relatively complex. This text is meant to serve as an introduction. Readers that are interested in additional coverage may want to review other excellent sources (Guo & Fraser, 2010, 2015). My objectives are that, by the end of this chapter, you will be able to (a) understand the basic concepts of PSA, (b) estimate propensity scores, (c) understand various conditioning strategies and be able to match units on their propensity scores, and (d) check model adequacy.
