I have measured the attribute importance of a food service using AHP. There were 42 pairwise comparisons.
In many responses (about 30%), all or almost all pairs were rated equally important. How should I handle these responses? Should I keep all responses or eliminate those with too many answers of equal importance? Your suggestions will be highly appreciated.
Dear Suranaree
Well, it could be that the responses reflect what people think, however, the high rate of coincidences or quasi coincident make the exercise of very dubious validity.
There could be another factor, such as the complexity of the questions or the diffcult for people to make pair-wise comparisons, and in reality, 42 were too many for my taste.
However, the most probable reason was that people got bored with so many questions and just answered anyrthing, just to get rid of it, being equality the most probable.
In my opinion, a big problem in pair-wise conparsion is that people have to compare things on asubjects on which they perhaps don't have any idea.
If you ask a guy, that considering an unique factor, to decide which is the relative importance of criterion colesterol with respect to criterion size of dishes, chances are that the person does not have the faintest idea of what you are asking.
In my opinion, the data you collected, using the right way, is useless.
Dear Nolberto Munier
Thank you very much for your opinion The questions pertain to the general characteristics of restaurants, such as food quality, ambiance, and price. The 42 pairwise comparisons cover attribute and sub-attribute comparisons at two levels.
This is the first time I've utilized AHP. Although I made each question brief and simple, I've discovered that collecting data presents some challenges. Is it a good approach if I decide to keep only the responses with sufficient consistency and not too many pairs of equal weight? I plan to collect more data to cover the lost data.
Dear Sunitiya (sorry for my confussion with your name)
If you want my opinion, I believe that it is better to ask each person, "In a 1 to 10 range, the higher the better, how do you rank quality of food" ?, and the same for each criterion. Compute an average for each criterion, put that value in the decision matrix, and use a MCDM method.
For prople it is much easier than having to make comparisons.
Forget about attributes and sub attributes, just treat each one independently.
I know that this porcedeure does not apply to AHP, but it does to most methods.
By the way AHP is deeply flawed.
You talk about consistency. Consistency on what?
The problem with elimminating criteria is that you probably will not be representing duly the problem. For instance, suppose that you decide to delete criteria 'prize' and 'quality'. In so doing you will be not using the two most important criteria.
Dear Nolberto Munier
In terms of consistency, I mean the AHP consistency ratio (Sorry for my unclear writing)
Thank you very much for your suggestion on the MCDM research method and your further thoughts. Your suggestion is very helpful.
Thank you Sunitiya
The consistency ratio is a factor to test the coherence of the DM judgements , and has no relation to the reality of the problem under study. It is absolutely useless to analyze a real situation, which may be correct for one DM and different for another. Thus, the AHP 'mathematial theory' convalidates that there are several realities for the same problem.
Until now, and insensible to my many requests to this respect, nobody has been able to explain me that WHY what is in the mind of the DM is applicable to the real world.
You are most welcome. I suggest again, to use a rational method. There are many, all have drawbacks of course, but none at the fantasy level of AHP
- Badges
- Science topic
- Similar topics
- Engineering
- Control Theory
- Control Systems
- Control Systems Engineering