Dear Bognar

Reference is made to your article:

“Development of the PRISM Risk Assessment Method Based on a Multiple AHP-TOPSIS Approach”

It is indeed an important article since risk is not a subject frequently addressed.

I read it, and found some aspects on what I am afraid don’t agree with you.

You use two very well known MCDM methods, AHP and TOPSIS. I don’t have any reserve for TOPSIS, one of the best and more coherente MCDM methods, but not for AHP, because its lack of rationality.

1- In my opinion, your paper appears to make a very common omission, I would say frequent in maybe 90 or 95 % of AHP users. It is not something attributable to the method, but to the DMs, and it is that this method, as you know, works only with independent criteria, as was clearly specified by Saaty.

I believe that the three components of risk, i.e., probability of occurrence, consequences and detection, are related, consequently, AHP should not be used here.

If you consider for instance, risk of earthquakes, probability of occurrence or frequency, can be estimated via statistics. It is evident that detectionis related to occurrence.

It is not the same an earthquake with an averaged frequency of 2 times a year, compared with earth tremors of different intensities several times a year, as happens for instance, in Los Angeles. Obviously, the higher the frequency the greater the probability, since Probability = Frequency / Possible outcomes.

Examining consequences to occurrence, it is not the same the consequences of an earthquake of 7 in the Richter scale occurring in a desert than in a populated area. In the first, the consequences are nil, but in the second they can be catastrophic.

2- You say: 'Development of the TOPSIS method by combining itwith AHP for solid factor weight determination and decision making’

Do you really believe that a selection of criteria importance built on the basis on DMs intuition is solid? Hard to believe.

RPN = Severity x Occurrence x Detection, represents a product of probabilities, which makes sense, because each one of these factors are based on statistics, knowledge, an experience, and these qualities are absent in AHP.

3- When talking about PRISM you mention that it works with partial risk. In my opinion, that is a simplification, since normally, a risk produces a series of consequences. A tsunami for instance, like the one in Fukushima, triggers a series of risks like:

Tsunami > Main transformers > Emergency transformers > Diesel Electric Generators > Batteries set > Seawater flooding

Where the ‘>’ symbol means ‘Precedence of failure’. Thus, a failure of the main transformers activates the emergency transformers, their failure activates the Diesel Electric Generators, and so on.

Therefore, we can’t consider the first risk only, in this case a tsunami, but the series of risks and their consequences. In addition, PRISM considers the same weight for each of the three dimensions, when clearly it may not happen.

4- I reckon that your proposal in combining the three factors for each incidence is, as far as I know, a new procedure, attractive, and that can be very useful, but I don’t think that pair-wise comparison using AHP is the most appropriate method, by producing arbitrary weights, and without any mathematical support. On the contrary, using TOPSIS is a very good approach, because its rationality that does not compare well with the irrationality of AHP.

Now, if the weights are determined individually, with reasoning, experience, knowledge, and research, they are are credible.

5- What called my attention is indeed a very important aspect, when you say that your method can uncover hidden risk. Sorry, I don’t realize how you reach that aspect.

Could you please explain it? It really interests me.

Thank you

Nolberto Munier