NotesFAQContact Us
Collection
Advanced
Search Tips
Peer reviewed Peer reviewed
PDF on ERIC Download full text
ERIC Number: ED562419
Record Type: Non-Journal
Publication Date: 2015
Pages: 10
Abstractor: ERIC
Reference Count: 14
ISBN: N/A
ISSN: N/A
Reducing Bias and Increasing Precision by Adding Either a Pretest Measure of the Study Outcome or a Nonequivalent Comparison Group to the Basic Regression Discontinuity Design: An Example from Education
Tang, Yang; Cook, Thomas D.; Kisbu-Sakarya, Yasemin
Society for Research on Educational Effectiveness
Regression discontinuity design (RD) has been widely used to produce reliable causal estimates. Researchers have validated the accuracy of RD design using within study comparisons (Cook, Shadish & Wong, 2008; Cook & Steiner, 2010; Shadish et al, 2011). Within study comparisons examines the validity of a quasi-experiment by comparing its estimates to trustworthy benchmarks (usually an experiment) with the same treatment group. First developed by Lalonde (1986), it is a rigorous method to evaluate the internal validity of quasi-experiments. Despite the accuracy of RD design and its increasing popularity, the basic RD design has three major disadvantages. The first disadvantage is that RD has weak statistical power. Previous literature (Goldberger, 1972; Schochet, 2009) shows that RD needs at least 2.75 times of RCT's sample size to achieve the same statistical power. The second disadvantage is that RD can only estimate causal effect at the cutoff, but most times researchers are concerned with average treatment effect across all populations. The third disadvantage is that RD depends on specification of functional forms. Researchers usually specify several different functional forms to test the sensitivity of estimates. To mitigate these three weaknesses, researchers have proposed many variants of RD (Lohr, 1972; Shadish, Cook and Campbell, 2002). Wing and Cook (2013) employed within study comparisons in the healthcare context and examined the advantages of adding a pretest to a basic RD, thus creating comparative regression discontinuity with the pretest as a comparison function (CRD-Pre). Figures are appended.
Society for Research on Educational Effectiveness. 2040 Sheridan Road, Evanston, IL 60208. Tel: 202-495-0920; Fax: 202-640-4401; e-mail: inquiries@sree.org; Web site: http://www.sree.org
Publication Type: Reports - Research
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: Society for Research on Educational Effectiveness (SREE)