Abstract: An International Comparison of Private and Public Schools Using Multilevel Propensity Score Methods and...

Area of Science:

• Statistics
• Educational Research
• Data Visualization

Background:

• Propensity score analysis (PSA) is increasingly used for causal inference in observational studies.
• Existing PSA methods have limitations with multilevel or clustered data, lacking specialized graphical tools.
• Visualizing propensity score analysis for multilevel data remains an underdeveloped area.

Purpose of the Study:

• Introduce the multilevelPSA R package for estimating propensity scores and visualizing results with multilevel data.
• Extend the graphical framework for propensity score analysis to accommodate multilevel data structures.
• Provide tools for nuanced understanding of causal effects in complex, hierarchical datasets.

Main Methods:

• Development and implementation of the multilevelPSA package in R.
• Application of cluster-based functions for propensity score estimation in multilevel data.
• Utilizing enhanced graphics to visualize propensity score analysis results for multilevel comparisons.

Main Results:

• The multilevelPSA package facilitates cluster-based propensity score estimation and visualization for multilevel data.
• Analysis of international Programme for International Student Assessment data revealed nuanced differences between private and public schools.
• Graphical representations provided a more detailed understanding of adjusted differences in educational outcomes across countries than statistical significance alone.

Conclusions:

• The multilevelPSA R package effectively extends propensity score analysis to multilevel data.
• Modern graphics offer enhanced insights into multilevel comparisons, complementing traditional statistical summaries.
• Visualizations derived from propensity score methods can be readily interpreted by diverse audiences, including non-technical ones.