Thread: normalizing reverse value of 0

I have the following problem

I have 3 options with different ranks:
A=3
B=0
C=0

I would like to normalize them to 0..1 scale.
The problem is that actually the value 0 is "better" than value 3, so I would reverse the values beforehand and use 1/A, 1/B and 1/C instead.

So the normalized weigh for A on the 0..1 scale would be:
(1/A)/((1/A)+(1/B)+(1/C))

For B:
(1/B)/((1/A)+(1/B)+(1/C))

For C:
(1/C)/((1/A)+(1/B)+(1/C))

However if A, B or C are zeros there can not be any reverse value. How to normalize in this case?

Thank you.

Originally Posted by cross777

I have the following problem

I have 3 options with different ranks:
A=3
B=0
C=0

I would like to normalize them to 0..1 scale.
The problem is that actually the value 0 is "better" than value 3, so I would reverse the values beforehand and use 1/A, 1/B and 1/C instead.

So the normalized weigh for A on the 0..1 scale would be:
(1/A)/((1/A)+(1/B)+(1/C))

For B:
(1/B)/((1/A)+(1/B)+(1/C))

For C:
(1/C)/((1/A)+(1/B)+(1/C))

However if A, B or C are zeros there can not be any reverse value. How to normalize in this case?

Thank you.

If you don't tell us what you want to do with these we cannot give any sensible suggestions.

CB

I want to use direct comparison to find which one is the best of the alternatives A, B or C.

And I would like to calculate normalized weights for this.


Logically A has probably normalized weight 0,
B has 0.5 and C has 0.5.
However, I am not able to write down the process how to calculate them