MULTICOLLINEARITY AND THE STATISTICAL POWER OF REGRESSION ANALYSIS,
Abstract
Multicollinearity is an econometric problem not connected with the error term and for that reason cannot be remedied by suitable manipulation of the data. One question, especially pertinent in light of the frequent use of stepwise regression programming routines, is the effect adding a collinear variable upon other coefficients and tests of significance. The question is particularly important in exploratory studies or studies in which little is known a priori about specification. In this type of study, tests of significance are more important than is usually the case, for they generally play a major role in determining the final specification. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1969
- Accession Number
- AD0681116
Entities
People
- Joseph P. Newhouse
Organizations
- RAND Corporation