Sensitivity analysis of additive weighted scoring methods – How to fool your friends (again)

PDF

Authors
  1. Kaluzny, B.L.
  2. Shaw, R.H.A.D.
Corporate Authors
Defence R&D Canada - Centre for Operational Research and Analysis, Ottawa ON (CAN);Directorate Materiel Group Operational Research, Ottawa Ont (CAN)
Abstract
This paper generalizes a sensitivity analysis approach originally suggested by Brereton [Brereton, R.C. (1977), Weighting Factors and How to Fool Your Friends, (Staff Note ORD SN 1977/06) Operational Research Division, Ottawa, Canada] for additive weighted scoring methods. Using high-dimensional computational geometry, three different sensitivity measure classes are proposed: volume, distance and representativity. The methodology can be used to analyze the subjective selection of criteria weights and provide decision makers with quantitative evidence to evaluate the robustness of the output ranking to small variations in the proposed weights. In contrast to Brereton’s approach, the proposed methodology can be used to analyze the sensitivity of a ranking for any subset of options. Furthermore the methodology is easy to apply using existing software available to analysts. The Acquisition Support Team (AST) applied the methodology on an options analysis problem faced by Project Management Office Canadian Surface Combatant (PMO CSC). AST demonstrated that the overall ranking of options presented by PMO CSC is stable against moderate variations in the weights and has not produced any evidence for fine-tuning.

Il y a un résumé en français ici.

Keywords
Support to Acq and LCC;Bid Evaluation;Borda Count;Multi Criteria Decision Analysis;Polytope;Ranking;Scoring Rule;Sensitivity Analysis;Weight Allocation
Report Number
DRDC-CORA-TR-2009-002 — Technical Report
Date of publication
01 Apr 2009
Number of Pages
60
DSTKIM No
CA032638
CANDIS No
531838
Format(s):
Electronic Document(PDF)

Permanent link

Document 1 of 1

Date modified: