In AHP (Analytical Hierarchical Process) using 9 criteria, when I use a a scale of relative importance (1-9), I am getting a higher lambda-max and hence an inconsistent Consistency Ratio (>.10) is always getting generated. Any suggestion to get generate consistent CR (
Dear Joshy
And why there should be a consistency ratio? Who said that?
Fortunately for AHP users, that lack of consitency is 'corrected' or 'altered' by a formula, it does not matter if the DM likes it or not.
I do not think it is realistic to accept that a formula forces thre DM to accept what he estimated is incorrect for the sake of transitivity. Where is the logic here?
In my opinion, the reason for asking consistency is related to the capacity of EV that demands it, albeit I don't think it matters when using GM
The demand for a consistency ratio is rooted on the fact that AHP declares that preferences are constant, one of the many false assumptions of the method, irrelevant of the alternatives and the circumstances. It was decided that way and it does not matter if this is true or not.
As an example consider this case: A DM works with alternatives A, B and C, regarding two criteria, Quality and Price, and he asserts that for him quality is always preferable to price. Comparing alternatives A and B, he chooses alternative A, with higher quality and prize than alternative B, that is, he is willing to pay more by no choosing alternative B, that has a little less quality and also lower price. In so doing, he prefers to lose some money but gain in quality by selecting alternative A by following his principle. Nothing wrong with this.
But he also compares A with alternative C, and finds that in selecting C he will be losing a little quality when compared to A, but also that he can save money selecting C, which has a considerable lower price, and he decides that this is preferable to the first selection.
Comparing B and C, the DM has not preferences; for him, both are equal, or B = C.
Therefore, A>B, A< C, B=C. Even when there is no consistency, is there anything wrong?
Obviosly not, because the DM is analizing quantities and prices, based on the respective alternatives values for both, and selects the best for him.
In this last circumstance, his preference changed and this is absolutely correct, but AHP ignores it,because for it preferences are constant, and thus forces A>C
That is, against this reasonable analysis, AHP determines that it must violate the DM's preference and thus, it does not represent reality.
Which is the root of this drawback of AHP?
The reason is that criteria weights are selected not considering the alternatives, as if it were possible! Just go to any example in real life, and you will see that criteria weights depend on the alternatives, not the other way around.
This is only one of the AHP flaws.
I would love to read your answer, comments, or rebuttals
- Badges
- Science topic
- Similar topics
- Vectorization