A MONTE CARLO STUDY OF LINEAR REGRESSION ASSUMPTIONS,
Abstract
The report demonstrates the effects on linear regression estimates of variations in (1) the distribution of the dependent variable, (2) distribution of the independent variable, and (3) intercorrelation between independent variables. The basic approach is to premise a population model to reflect some underlying physical law or structural relationship, assign numerical values to parameters, and then introduce violations of the classical assumptions that might be typical in cost estimation problems. From the various cases thus constructed, samples are generated by computer in Monte Carlo fashion. This empirical study tests the sensitivity (in terms of bias and sampling variability) of regression coefficients and common statistical measures of reliability.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1969
- Accession Number
- AD0698734
Entities
People
- G. C. Sumner
Organizations
- RAND Corporation