De Leeuw, Jan
Dr. Armor's Regressions
University of California Los Angeles
Critique of statistical analyses used in desegregation research by Dr. Armor, as well as by well-known expert witnesses and researchers in desegregation litigation.
Journal Name or Institutional Affiliation:
UCLA Department of Statistics
UCLA Statistical Series 248
- DeLeeuw critiques the statistical analyses in Armor's research on desegregation and Black-White achievement.
- Criticizes Armor's "gap-analysis" and flawed interpretation of results when correcting for the covariates of the "Black" variable.
- Regression does not "explain" but establishes correlations between variables.
- Causal attribution as Armor concludes is not possible.
- Problems with Armor's labeling---treats gender and race as proxies for SES.
- Gap analysis procedure is unsound. Armor does not accurately assess the strengths and weaknesses of HLM and OLS.
- Armor overestimates the use, applicability, and stability of OLS.
- Armor's analytical method artificially suppresses the regression coefficient for school level variables, makes errors that are well documented in the stats literature as fallacies of OLS ("Robinson Effect").
- Author is internationally renown statistician and chair of UCLA Department of Statistics
Academic Achievement, Achievement Gap, Desegregation, SES
Method of Analysis:
Unit of Analysis:
- Critique of David Armor's previous studies used in desegregation cases.
- This article is a methodological critique of Armor's stepwise regression analyses in various desegregation cases.