Predictive Modeling With Missing Data: Monte Carlo Simulation With Social Science Panel Data

• In Event: 77.056 - Fairmont Royal York Roundtable Session Seven
  In Roundtable Session: 77.056-3 - Innovative Statistical Modeling and Methods

Abstract

Predictive modeling has begun to gain attention in social science, following the practice of machine learning. A body of simulation studies on predictive modeling with missing data in other fields such as computer science and medicine does not properly address the issues of social science data. As one of the first Monte Carlo simulation studies to investigate predictive modeling with missing data in social science, this simulation study endeavored to emulate a real social science panel data. The simulation conditions included missing data techniques (listwise deletion, k-NN, EM), missing rates (small, large), missingness mechanisms (MAR, MNAR), and penalized regression methods (LASSO, adaptive LASSO, and MCP). Implications to predictive modeling using incomplete social science panel data were discussed accordingly.

Authors

• Jin Eun Yoo, Korea National University of Education

• Minjeong Rho, Korea National University of Education