Propensity Score Matching on Discrete Treatment: Beijing Pm2.5 Case Study

Abstract—In causal inference, propensity score matching (PSM) is an effective method to estimate the causal effect between treatment and potential outcomes. The PSM with binary treatment has been widely used in medicine, economics, and sociology fields to evaluate the influence of treatment on the potential outcomes. However, the binary treatment is a special case of discrete treatment. The multi-level treatment is also a universal case of discrete treatment. Therefore, this essay will focus on the discrete treatment (from binary to multi-level) effect estimation by the propensity score matching method. In the procedure of propensity score matching, apart from the logistic model, more other machine learning models can be applied to estimate the propensity score for different types of treatment. This paper aims to combine machine learning models with propensity score matching and apply the methods to the Beijing pm 2.5 dataset.