CRITIC
AvailableCRiteria Importance Through Intercriteria Correlation. Diakoulaki et al. (1995)
The idea
Show as table
| Criterion | Contrast (σ) | Conflict Σ(1 − ρ) | Information content C | Weight |
|---|---|---|---|---|
| Price | 0.4062 | 4.3419 | 1.7637 | 0.1872 |
| Screen size | 0.4147 | 4.1752 | 1.7316 | 0.1838 |
| Pixel density | 0.4394 | 3.6254 | 1.5931 | 0.1691 |
| Thickness | 0.4161 | 5.8868 | 2.4494 | 0.2599 |
| Mass | 0.4063 | 4.6397 | 1.8850 | 0.2000 |
Data: Smartphones (Krishnan et al. 2021).
CRITIC gives more weight to criteria whose values vary a lot (contrast) and disagree with the other criteria (conflict).
When to use it
- Weights must come from the data, not from people.
- Correlated criteria (financial ratios, specifications) should weigh less.
- You want steps that are easy to show: deviation, correlation, information.
Algorithm
1Min-max normalization
J+ are benefit and J- cost criteria; the origin paper's "ideal point" transformation maps every column to 0-1, with 1 the best value.
2Contrast: standard deviation
Sample standard deviation (Excel STDEV); using m instead of m - 1 scales every sigma equally, so the weights do not change.
3Correlation between criteria
Pearson correlation of the normalized columns, without absolute value.
4Information content
Contrast times conflict: a criterion that is spread out and disagrees with the others carries more information.
5Weights
Used together with
Why these pairs
- CRITIC then TOPSIS is the pipeline this site started from. Keshavarz-Ghorabaee et al. (2021) review a material-selection study that feeds CRITIC weights into TOPSIS, GRA and ORESTE.
- Often run next to Entropy, SD and MEREC as competing objective weightings on the same matrix (Keshavarz-Ghorabaee et al. 2021, Table 6; Mukhametzyanov 2021).
- CRITIC then EDAS appears in Turkish financial-performance studies, for example Bayram (2021) on participation banks.
Reference example
Details
Krishnan, A.R.; Kasim, M.M.; Hamid, R.; Ghazali, M.F. (2021). A Modified CRITIC Method to Estimate the Objective Weights of Decision Criteria. Symmetry 13(6), 973. doi:10.3390/sym13060973
Table 1 (input), Table 2 and Table 5 (output). Five smartphones and five criteria. Our weights match the published ones to 4 decimals (largest difference 4.6e-5) and the standard deviations match Table 2. A second check against the project's own Excel sheet agrees to 1e-12.
Checked by this site's test suite.
Sources
Show 5 sources
- Diakoulaki et al. (1995). Determining objective weights in multiple criteria problems: The CRITIC method. Computers & Operations Research 22(7), 763-770. doi:10.1016/0305-0548(94)00059-HThe steps follow Diakoulaki et al. (1995) directly and agree with the open-access restatements by Krishnan et al. (2021) and Mukhametzyanov (2021).
- Diakoulaki, Mavrotas & Papayannakis (1995), Computers & Operations Research 22(7), 763-770 doi:10.1016/0305-0548(94)00059-H
- Krishnan, Kasim, Hamid & Ghazali (2021), Symmetry 13(6), 973 doi:10.3390/sym13060973
- Mukhametzyanov (2021), Decision Making: Applications in Management and Engineering 4(2), 76-105 doi:10.31181/dmame210402076i
- Keshavarz-Ghorabaee et al. (2021), Symmetry 13(4), 525 doi:10.3390/sym13040525