Propensity score matching

Propensity Score Matching

Propensity Score Matching (PSM) is a statistical technique used in observational studies to estimate the effect of a treatment, policy, or other intervention by accounting for the covariates that predict receiving the treatment. It is a method of reducing selection bias by equating groups based on these covariates.

Overview

The propensity score is the probability of a unit (e.g., a person, a school) being assigned to a particular treatment given a set of observed covariates. The concept was introduced by Paul Rosenbaum and Donald Rubin in 1983. The main idea is to balance the distribution of observed covariates between treated and control groups, thereby mimicking some of the characteristics of a randomized controlled trial.

Methodology

The process of Propensity Score Matching involves several steps:

1. **Modeling the Propensity Score**: The first step is to estimate the propensity score using a statistical model, typically a logistic regression, where the treatment assignment is regressed on observed covariates.

2. **Matching**: Once the propensity scores are estimated, units in the treatment group are matched with units in the control group that have similar propensity scores. Common matching methods include nearest neighbor matching, caliper matching, and kernel matching.

3. **Assessing Balance**: After matching, it is crucial to assess whether the covariates are balanced between the treatment and control groups. This can be done using standardized mean differences or other balance diagnostics.

4. **Estimating Treatment Effects**: Finally, the treatment effect can be estimated by comparing outcomes between the matched treatment and control groups. This can be done using simple difference in means or more complex models.

Advantages and Limitations

Propensity Score Matching offers several advantages:

- It reduces selection bias by balancing covariates between treated and control groups. - It allows for causal inference in observational studies where randomization is not possible.

However, there are also limitations:

- It only accounts for observed covariates, so unobserved confounders can still bias the results. - The quality of the matching depends on the correct specification of the propensity score model.

Applications

PSM is widely used in various fields such as epidemiology, economics, education, and social sciences. It is particularly useful in policy evaluation and health services research where randomized controlled trials are not feasible.

Related Pages

• Causal Inference
• Randomized Controlled Trial
• Selection Bias
• Observational Study
• Covariate