Scribd
Upload
91 views

Annexure 2

The document summarizes the Analytic Hierarchy Process (AHP) theory and calculations for determining the relative importance or weight of elements being compared. It explains that AHP involves forming a reciprocal matrix of judgments, finding the eigenvector of the matrix to represent the weights, and calculating a Consistency Ratio to assess the consistency of the judgments. It then provides a worked example comparing four attributes (A, B, C, D) with the eigenvector showing C is most important and A is least, and calculates the Consistency Ratio of 0.0677 which is below the 0.1 threshold indicating consistent judgments.

Uploaded by

Sanchari_rittika
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
91 views

Annexure 2

The document summarizes the Analytic Hierarchy Process (AHP) theory and calculations for determining the relative importance or weight of elements being compared. It explains that AHP involves forming a reciprocal matrix of judgments, finding the eigenvector of the matrix to represent the weights, and calculating a Consistency Ratio to assess the consistency of the judgments. It then provides a worked example comparing four attributes (A, B, C, D) with the eigenvector showing C is most important and A is least, and calculates the Consistency Ratio of 0.0677 which is below the 0.1 threshold indicating consistent judgments.

Uploaded by

Sanchari_rittika
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

ANNEXURE 2 THE AHP THEORY AND CALCULATIONS The mathematical basis of the AHP can be explained in fairly simple

outline for the purposes of this book but one needs to know what a matrix and a vector are and how to multiply a matrix by a vector. We will cover the mathematics first and then explain the calculations. THE AHP THEORY Consider n elements to be compared, C1 Cn and denote the relative weight (or priority or significance) of Ci with respect to Cj by aij and form a square matrix A=(aij) of order n with the constraints that aij = 1/aji, for i j, and aii = 1, all i. Such a matrix is said to be a reciprocal matrix. The weights are consistent if they are transitive, that is aik = aijajk for all i, j, and k. Such a matrix might exist if the aij are calculated from exactly measured data. Then find a vector of order n such that A = . For such a matrix, is said to be an eigenvector (of order n) and is an eigenvalue. For a consistent matrix, = n. For matrices involving human judgement, the condition aik = aijajk does not hold as human judgements are inconsistent to a greater or lesser degree. In such a case the vector satisfies the equation A= max and max n. The difference, if any, between max and n is an indication of the inconsistency of the judgements. If max = n then the judgements have turned out to be consistent. Finally, a Consistency Index can be calculated from (max-n)/(n-1). That needs to be assessed against judgments made completely at random and Saaty has calculated large samples of random matrices of increasing order and the Consistency Indices of those matrices. A true Consistency Ratio is calculated by dividing the Consistency Index for the set of judgments by the Index for the corresponding random matrix. Saaty suggests that if that ratio exceeds 0.1 the set of judgments may be too inconsistent to be reliable. In practice, CRs of more than 0.1 sometimes have to be accepted. A CR of 0 means that the judgements are perfectly consistent. THE AHP CALCULATIONS There are several methods for calculating the eigenvector. Multiplying together the entries in each row of the matrix and then taking the nth root of that product gives a very good approximation to the correct answer. The nth roots are summed and that sum is used to normalise the eigenvector elements to add to 1.00. In the matrix below, the 4th root for the first row is 0.293 and that is divided by 5.024 to give 0.058 as the first element in the eigenvector. The table below gives a worked example in terms of four attributes to be compared which, for simplicity, we refer to as A, B, C, and D.

The eigenvector of the relative importance or value of A, B, C and D is (0.058,0.262,0.454,0.226). Thus, C is the most valuable, B and D are behind, but roughly equal and A is very much less significant. The next stage is to calculate max so as to lead to the Consistency Index and the Consistency Ratio. We first multiply on the right the matrix of judgements by the eigenvector, obtaining a new vector. The calculation for the first row in the matrix is: 1*0.058+1/3*0.262+1/9*0.454+1/5*0.226 = 0.240 and the remaining three rows give 1.116, 1.916 and 0.928. This vector of four elements (0.240,1.116,1.916,0.928) is, of course, the product A and the AHP theory says that A=max so we can now get four estimates of max by the simple expedient of dividing each component of (0.240,1.116,1.916,0.928) by the corresponding eigenvector element. This gives 0.240/0.058=4.137 together with 4.259, 4.22 and 4.11. The mean of these values is 4.18 and that is our estimate for max. If any of the estimates for max turns out to be less than n, or 4 in this case, there has been an error in the calculation, which is a useful sanity check. The Consistency Index for a matrix is calculated from (max-n)/(n-1) and, since n=4 for this matrix, the CI is 0.060. The final step is to calculate the Consistency Ratio for this set of judgements using the CI for the corresponding value from large samples of matrices of purely random judgments using the table below, derived from Saatys book, in which the upper row is the order of the random matrix, and the lower is the corresponding index of consistency for random judgements.

For this example, that gives 0.060/0.90=0.0677. Saaty argues that a CR > 0.1 indicates that the judgements are at the limit of consistency though CRs > 0.1 (but not too much more) have to be accepted sometimes. In this instance, we are on safe ground. A CR as high as, say, 0.9 would mean that the pairwise judgements are just about random and are completely untrustworthy.

You might also like