Why Regression Coefficients Have the Wrong Sign
- 1 July 1976
- journal article
- research article
- Published by Taylor & Francis in Journal of Quality Technology
- Vol. 8 (3) , 121-126
- https://doi.org/10.1080/00224065.1976.11980732
Abstract
It is not uncommon that regression coefficients appear with the sign opposite to the expectations of the researcher. Several causes and cures of this data analyst's nightmare are gathered together in this paper. Among the causes discussed are: range of independent variables is too small, excluding important variables from the model, multicollinearity and computation error. Some original results are presented and literature results are surveyed.Keywords
This publication has 8 references indexed in Scilit:
- Regression analysis and problems of multicollinearityCommunications in Statistics, 1975
- Some Aspects of Nonorthogonal Data AnalysisJournal of Quality Technology, 1973
- A New Method for Examining Rounding Error in Least-Squares Regression Computer ProgramsJournal of the American Statistical Association, 1971
- Fitting Equations to Mixture Data with Restraints on CompositionsJournal of Quality Technology, 1970
- Generalized Inverses, Ridge Regression, Biased Linear Estimation, and Nonlinear EstimationTechnometrics, 1970
- A Report on the Accuracy of Some Widely Used Least Squares Computer ProgramsJournal of the American Statistical Association, 1970
- An Appraisal of Least Squares Programs for the Electronic Computer from the Point of View of the UserJournal of the American Statistical Association, 1967
- Multicollinearity in Regression Analysis: The Problem RevisitedThe Review of Economics and Statistics, 1967