Simulate RD analysis files by selectively dropping observations from experimental data files¶

Strategy · All strategies
Evidence · no claims cited

Description¶

An implementable methodological recipe for creating RD analysis data from experimental data: "We simulate RD analysis files by selectively dropping observations from the original experimental data files." Observations on one side of the assignment variable's cutoff are dropped so that remaining units resemble an RD sample, and RD impact estimates can then be compared with the original experimental estimates. The article also recommends simulating manipulation by letting some units alter their assignment-variable values to receive treatment.

Design Implications¶

Context¶

Requirements¶

• An original experimental data set with an assignment variable and outcome measures for the same intervention.

Constraints¶

• The approach was demonstrated on two evaluations of educational interventions; the article does not claim generalizability beyond that setting.

Target Learners¶

• methodologists and evaluators of educational programs

Target Learning Goals¶

• assessing the validity of quasi-experimental impact estimates

Related Strategies¶

Examples¶

-

Key Sources¶

• Philip Gleason, Alexandra Resch, Jillian Berk. (2018). RD or Not RD: Using Experimental Studies to Assess the Performance of the Regression Discontinuity Approach. Evaluation Review, vol. 42. https://www.mathematica.org/publications/rd-or-not-rd-using-experimental-studies-to-assess-the-performance-of-the-regression-discontinuity