Two Enhancements of the Logarithmic Least-Squares Method for Analyzing Subjective Comparisons
Abstract
This memorandum presents two extension of the logarithmic least- squares method for analyzing subjective comparisons. The first is a test of consistency that is based on analysis-of-variance techniques. Unlike Saaty's eigenvalue-based consistency ratio, the consistency metric developed here is independent of any scale used to elicit comparisons. It is easily applied to sets of comparisons that are incomplete or contain multiple comparisons. This metric provides valuable insights into the trade-off between data set completeness and our ability to detect inconsistency. The second extension is to use robust regression for calculating the ratio scale that best represents most of the data. These extensions combine to make the logarithmic least-squares method an attractive alternative to Saaty's eigenvalue method. Keywords: Test of normality; Statistical tests.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 25, 1989
- Accession Number
- ADA211192
Entities
People
- Michael L. Mccurdy