Currently, there are issues with the implementation of MSM refinement methods. The pooled data set used to generate the propensity scores should contain data from only the time of treatment, not everything from t:t+f. Within each iteration of finding propensity scores and then weights before multiplication, the calculations should then almost exactly mirror the "non-MSM" methods, the major difference being the multiple iterations and multiplication.
Upon further review, I'm realizing what I had previously is not correct. What is the updated process/formula for calculating the weights?
adamrauh
commented
5 years ago
Currently, there are issues with the implementation of MSM refinement methods. The pooled data set used to generate the propensity scores should contain data from only the time of treatment, not everything from t:t+f. Within each iteration of finding propensity scores and then weights before multiplication, the calculations should then almost exactly mirror the "non-MSM" methods, the major difference being the multiple iterations and multiplication.Upon further review, I'm realizing what I had previously is not correct. What is the updated process/formula for calculating the weights?