Question

Variables that are affected by the treatment are generally not be included in the propensity score. Why do you think this is so? Can you think of an example?

Answer 1

The propensity score is used to ensure that the trial is performed in a randomized manner. This is essential to ensure that the treatment and control groups are comparable. If there is an inherent bias in the treatment group, the results that are obtained by the study will be biased and hence unreliable.

Therefore, the variables that are affected by the treatment group are generally not accepted in the propensity score. This is because such variables will have differential effects on the treatment and the control groups.

An instance of this is when an experiment is done to check for the effect of smoking on some parameters. Here the treatment and control groups cannot be randomly set up as making people smoke isn't ethical. However, smokers will have a bias towards a particular sex and age group. Hence these variables shouldn't be included in the propensity score.