The performance of propensity score matching and weighting methods on samples of various properties

Man, Qidi

Permalink

https://hdl.handle.net/2142/121404 Copy

Description

Title

The performance of propensity score matching and weighting methods on samples of various properties

Author(s)

Man, Qidi

Issue Date

2023-06-01

Director of Research (if dissertation) or Advisor (if thesis)

Anderson, Carolyn

Committee Member(s)

• Kern, Justin
• Jiang, Ge

Department of Study

Educational Psychology

Discipline

Educational Psychology

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• propensity score
• matching
• weighting
• observational study

Abstract

Observational studies tend to have samples with selection bias and therefore confound the treatment effect by variables that can influence both treatment and outcome, and propensity score analysis has become more widely used to process data in these conditions. Various types of propensity score matching or weighting methods can be used to balance covariate distributions between treatment and control groups and simulate randomized experiments, including matching methods like nearest neighbor matching without replacement, matching with replacement, optimal matching, variable ratio matching, and weighting methods like inverse probability of weighting and overlap weighting. This study compares effect estimation performance in terms of bias and mean square error (MSE) after balancing covariates of simulated observational data using matching or weighting methods in different propensity score distribution conditions. The study also examines the effect of using different variable selection strategies and various estimators for effect estimation. The results suggest that the propensity score method choice and variable selection strategies can minimize the MSE of causal effect estimation in different scenarios. This study complements previous studies by providing guidance for the practical use of these propensity score-based methods. Limitations and future directions are also discussed.

Graduation Semester

2023-08

Type of Resource

Thesis

Copyright and License Information

Copyright 2023 Qidi Man

Owning Collections