Reciprocal matrices with random coefficients (Q3323112)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 3854274
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Reciprocal matrices with random coefficients |
scientific article; zbMATH DE number 3854274 |
Statements
Reciprocal matrices with random coefficients (English)
0 references
1982
0 references
matrix of preferences
0 references
largest eigenvalue
0 references
principal right eigenvector
0 references
gamma distributed random variables
0 references
Dirichlet random variable
0 references
empirical preference matrix
0 references
consistency
0 references
0.89120936
0 references
0 references
0 references
0.88050115
0 references
Let \(w=[w_ 1,w_ 2...w_ n]\), \(\Sigma_ iw_ i=1\), \(w_ i>0\), denote the priorities for n objects, and \(X=[X_{ij}=1/X_{ji}=w_ i/w_ j]\) the (reciprocal) matrix of preferences. The structure of X is X\=w\(=\lambda_{\max}\bar w\), where \(\lambda_{\max}=n\) is the largest eigenvalue and \(\bar w\) is the principal right eigenvector. If \(X_{ij}\sim \Gamma(\alpha_ i,\beta_ i)\) are gamma distributed random variables, then the right principal eigenvector is a Dirichlet random variable, \(\bar w\sim D(\alpha_ 1,\alpha_ 2...\alpha_ n)\), \(EX_{ij}=\alpha_ i/\alpha_ j\), and \(Ew_ i=\alpha_ i\times cons\tan t.\) An empirical preference matrix as elicited from pairwise comparison judgements from decision makers or experts is most likely not to be consistent, \(X_{ij}X_{jk}\neq X_{ik}\) for some i,j,k; one has to consider n(n-1)/2 independent random variables and the stochastic system of equations \(\Sigma_ jX_{ij}w_ j=\Lambda_{\max}w_ i\). Based on 500 randomly generated matrices X, the distribution of \(\Lambda_{\max}\) is obtained, and a test is constructed which allows to decide wether X can be treated as consistent, i.e. \(\hat X_{ij}\hat X_{jk}=\hat X_{ik}\) and \(\hat w\sim \hat D({\hat \alpha})\). If the relaxed assumption of consistency is tenable, the principal right eigenvector is an estimate for the priorities.
0 references