View Single Post
  #8   Report Post  
Posted to microsoft.public.excel.worksheet.functions
Aaron Aaron is offline
external usenet poster
 
Posts: 59
Default Auto range of numbers detection

On Feb 24, 4:34 am, JE McGimpsey wrote:
In article . com,

"Aaron" wrote:
Another way to look at it is, if you plot the data on a bar graph, the
groupings are easier to see. You will be able to form a trend between
contiguous numbers and make "groups" out of the numbers accordingly.


A bar graph may make it more obvious to you, but it doesn't address the
fundamental problem.

Both XL and VBA need specific, explicit criteria in order to calculate.
There's no fuzzy logic engine built-in, which is what you're asking for.

I do this in my head.


And that's the issue - what you do in your head is obvious to you, but
not obvious at all to a designer of an algorithm. One would need dozens
of examples, encompassing the entire range of possible values, along
with your judgment on how to group before one could design the kind of
fuzzy logic analogue that you're using in your head.

For example, based on the examples you've given it's entirely unclear to
me, even if plotted as a bar graph, what rules to apply to decide
whether this is one group or three:

0.001, 0.001, 0.002, 0.002, 0.001, 0.001

or whether this is two groups, or four, or six:

0.001, 20, 0.002, 525, 556.001, 524


Yes I suppose this is true, I was hoping that there might just be a
formula or way of linking formulas that would be able to do what I
would like to do.

The errors that you have shown on the first example are typical of
what I work with, but the second example would never happen.

The calculations that lead me to the number in the 10 excel columns is
a simple a minus b formula. (Error minus what it should be.)

The errors therefore will be close to each other to start with and on
a perfect calibration of equipment, there will be no errors at all.

The information I know that could apply to all examples is stuff like:

the exact amount of decimal places it will be,
that they will all be the same decimal place,
that the minimum division size in the error will be in a multiple of
0.1 or 0.5 for example,

0.1
0.2
0.3
0.1
or

0.5
0.15
0.10

etc not a mixture,

And that the errors I get could be either positive or negative.

For the examples you stated above, I would group the first 2, the
second 2 and the last 2, making 3 groups.

the second example would be 4 groups, the first number would be one,
the second and the third would be another and the last 3 all in their
own group.

If you wrote them this way, 0.001, 0.002, 20, 525, 556.001, 524 then I
would group the first 2 in one group, the third on its own and the
last 3 together.

Cheers,

Aaron.