Merging Ranked Subsets of Performance Appraisals.
Abstract
Rank ordering the elements of a set too large for one rater to survey can be accomplished by having numerous raters rank overlapping or intersecting subsets of this population. The least squares model of Gulliksen for evaluating incomplete paired comparison data is extended for use with weights and ranking data. An analysis of variance model is developed for analyzing agreement between raters, and the accuracy of the least squares solution. Minimization of the L sub 2 norm of errors is compared with minimization of the L sub 1 norm determined using linear programming. The models are compared using a computer simulation. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1972
- Accession Number
- AD0757811
Entities
People
- Spencer Bryce Graves
Organizations
- University of Pittsburgh