ROC
AvailableRank-order weights: Rank Order Centroid, Rank Sum, Rank Reciprocal. Barron and Barrett (1996)
The idea
Circles mark ROC (simplex centroid), squares mark RS (linear) and diamonds mark RR (reciprocal) on the same scale; the bars on the right show the active formula's weights.
Show as table
| Criterion | Rank r | ROC | RS | RR | Weight w |
|---|---|---|---|---|---|
| C1 | 1 | 0.4567 | 0.3333 | 0.4380 | 0.4567 |
| C2 | 2 | 0.2567 | 0.2667 | 0.2190 | 0.2567 |
| C3 | 3 | 0.1567 | 0.2000 | 0.1460 | 0.1567 |
| C4 | 4 | 0.0900 | 0.1333 | 0.1095 | 0.0900 |
| C5 | 5 | 0.0400 | 0.0667 | 0.0876 | 0.0400 |
Data: Five alternatives, five ranked criteria (Roszkowska 2013).
ROC, Rank Sum and Rank Reciprocal turn a ranking of criteria into weights.
When to use it
- Only ordinal information is available or agreed on (groups, time pressure).
Algorithm
1Rank order centroid (default)
r is the rank of the criterion (1 = most important), and each formula sums to 1 without further normalization, with w_j = w(r_j).
2Rank sum
3Rank reciprocal
Used together with
Why these pairs
- ROC, RS or RR then SAW, and a comparison of AHP with rank weights: Roszkowska (2013).
- Rank weights as surrogate weights in decision-quality simulations and compared against AHP pairwise weights: Barron & Barrett (1996); Roszkowska (2013).
Reference example
Details
Roszkowska, E. (2013). Rank Ordering Criteria Weighting Methods: a Comparative Overview. Optimum. Studia Ekonomiczne 5(65), 14-33. doi:10.15290/ose.2013.05.65.02
Published example (Table 4, row n = 4, and Tables 5-6, n = 5). For n = 4, ROC, RS and RR match the printed 2 decimals (0.52, 0.27, 0.15, 0.06; 0.40, 0.30, 0.20, 0.10; 0.48, 0.24, 0.16, 0.12). In the 5 x 5 case study (Tables 5-6), ROC weights give the same SAW ranking A4 > A2 > A5 > A1 > A3.
Checked by this site's test suite.
Sources
Show 6 sources
- Barron and Barrett (1996). Decision Quality Using Ranked Attribute Weights. Management Science 42(11), 1515-1523. doi:10.1287/mnsc.42.11.1515Roszkowska (2013) also credits Solymosi & Dombi (1986) for centroid weights, and Rank Sum and Rank Reciprocal come from Stillwell, Seaver & Edwards (1981); the formulas follow Roszkowska (2013).
- Barron & Barrett (1996), Management Science 42(11), 1515-1523 doi:10.1287/mnsc.42.11.1515
- Stillwell, Seaver & Edwards (1981), Organizational Behavior and Human Performance 28(1), 62-77 doi:10.1016/0030-5073(81)90015-5
- Solymosi & Dombi (1986), European Journal of Operational Research 26(1), 35-41 doi:10.1016/0377-2217(86)90157-8
- Roszkowska (2013), Optimum. Studia Ekonomiczne 5(65), 14-33 doi:10.15290/ose.2013.05.65.02
- pyDecision 5.1.1, roc_method and rsw_method github.com/Valdecy/pyDecision