Multiple Imputation With Artificial Intelligence: Missing Data in Propensity Score Analysis (Poster 6)

• In Event: AERA Poster Session Five
  In Poster Session: Division D Section 2 Poster Session

Abstract

Propensity score analysis may result in biased estimates if researchers do not handle missing data correctly. Traditional strategies for handling missing data may introduce bias, lack of power, and sensitivity to different data types. Our study introduces researchers to imputation with artificial intelligence (AI) to address these weaknesses. We compare AI to existing methods in a Monte Carlo experiment concerning the covariate balance using standardized mean differences measures, mean squared error of treatment effects, Type I error rates, and efficiency. Additionally, we apply AI imputation in analyzing the educational and employment experiences of street-identified minority men and women between the ages of 18–35.

Authors

• Zachary K. Collier, University of Connecticut
  Presenting Author

• Kamal Chawla, University of Delaware
  Non-Presenting Author

• Olushola Soyoye, University of Delaware
  Non-Presenting Author