Toward the Extraction of Educational Evidence: Proposal and Discussion of a Method for Adjustment of Background Information Difference

Abstract

Non-randomized controlled trials (non-RCTs) are easier to conduct than randomized controlled trials (RCTs), but differences in background information such as sex ratio and performance levels between the groups being compared are problematic. In this study, data from a previous non-RCT were used and adjusted using propensity scores. The differences between groups were reduced. The results suggest that non-RCTs can be made to resemble RCTs by using causal inference, and that high-quality evidence can be extracted in an easy way.