De Leeuw, Jan, Lew, Vivian, Bentow, Stan, & McKeever, Matt
Regression Analysis in the Wilmington Case
University of California at Los Angeles
Discusses the shortcomings of stepwise regression analyses in educational studies and examines the application of regression analyses by expert witnesses.
Journal Name or Institutional Affiliation:
UCLA Department of Statistics
UCLA Statistical Series 175
- He points out methodological flaws and problematic causal inferences drawn from their research.
- Major flaws in correcting for SES without accounting for the correction of race because of the high correlation between race and SES.
- Analysis is biased because it attributes most variation to SES and only the residual to race, rather than investigating the reverse.
- De Leeuw presents alternative regression analyses to the analyses in the reports of Armor, Achilles and Reschly.
- In contrast to Armor's and Reschly's findings, de Leeuw finds that Race Gap remains substantial after correcting for SES.
Achievement Gap, Desegregation, SES
Method of Analysis:
Unit of Analysis:
- De Leeuw defines regression analysis and discusses the uses, strengths, and common errors associated with regression analysis. He uses this information to review and critique expert reports written for the defense by Armor, Achilles, Walberg, Rossel, Raffel, and Reschly.
- Data analyses used in educational studies.
- Researchers served as expert witnesses in desegregation court cases.