search

insongkim / PanelMatch

113 stars 34 forks source link

multivalued treatment variables; survival outcomes #83

Open yatefi opened 3 years ago

yatefi commented 3 years ago

Hi, I had a number of questions regarding the package:

  1. I was curious if PanelMatch requires a binary treatment variable or can accommodate multivalued treatment effects as well (some of the matching methods such as CBPS used in the package can be used for multivalued treatment effects). What is the best strategy to use PanelMatch for multivalued treatment variables (treatment variables with more than 2 levels)?
  2. Can PanelMatch be used for survival analysis? If not directly in the package, is there a way to extract the calculated weights and use them as sampling weights in a separate survival analysis?
  3. Since in each time period a separate matching/weighting is performed, how should we use the weights in separate analyses? Can the weights be extracted?
  4. Traditionally with other propensity score weighting methods, you can extract the calculated weights and use them as regression weights. However, the calculated weights will typically be stored in 1 variable. Does PanelMatch give one variable as the weights or a matrix of variables (since it's done in each period)?