Part 1:
Write a proposal for a study using observational data that requires the use of propensity score analysis. Have two treatments defined by behaviors/actions of people. Keep in mind what covariates that explain or are related to that choice. These covariates are potential confounders. Other things to keep in mind: How long is the follow up in the study? What response variables will you choose to assess the differences between the treatments? Note that some covariates can be obtained with lab results (ie, baseline cholesterol levels in a study involving a cardiovascular disease)
Part 2:
Do an analysis with the "caith" data set (can be found in the package MASS). Packages vcd, ca, and MASS in R are all good options for the analysis.
(Part 2 can be expanded on)