Propensity Score Matching - How Can We Compare Apples with Peaches?

Zhou, Xinchang

Content Files

zhou_xinchang.jpeg

Permalink

https://hdl.handle.net/2142/126448

Description

Title

Propensity Score Matching - How Can We Compare Apples with Peaches?

Author(s)

Zhou, Xinchang

Contributor(s)

Yuhui Fu

Issue Date

2024

Keyword(s)

Educational Psychology

Abstract

Using randomized controlled trials for every situation isn't always feasible. For instance, we can't randomly assign some kids a new teaching method while leaving others out—it's up to students to decide whether to adopt it. The challenge arises because in non-random assignments, confounding factors like students’ skill level or preferred learning style can influence results. So, how do we fairly assess if the new method is effective?Propensity score is a predictive score indicating a student's likelihood to choose the new teaching method based on their background (e.g., skills, previous experience). Propensity score matching helps us pair students in/not in the new method class who share similar background characteristics. Then when compare results, we can be more confident that any differences are likely due to the teaching method, not other factors.In the picture, I showed the general idea of propensity score matching - finding similar individuals based on propensity scores and forming a new "matched sample". I also showed a part of my research results - when the data quality is different, various matching methods will bring different biases in treatment effect measurement.

Type of Resource

text
image

Language

eng

Owning Collections

Image of Research - Graduate Students PRIMARY