100% quandry

[Mojo]

Hi All

This might be an age old problem, but I just wanted to see if I wasn't
missing something obvious in my code.

Basically I have say 31 people (this figure can range from 10 to 100) who
are put into 8 categories. Once they have been categorised I simply need to
show how many are in each category as a number and as a percentage. Now the
number aspect of display is easy, but I'm coming unstuck with the percetange
aspect.

In essence as I do my % calc for each category sometimes certain values are
1% less or more than they should be and as you add up the displayed figures
the total might come out at 99% or 101%, which looks daft.

An example of a display problem is as follows

1 0 2 9 11 5 0 2 = 30 people
3% 0% 7% 30% 37% 17% 0% 7% = 101% !!!!

My code for each percentage value is simply:

Round((intPeopleCount / intTotalPeopleCount) * 100, 0) & "%"

I really want to make this work, but it seems fraught with issues!!

Any ideas?

Thanks

 

[Jim Cone]

answered in public.excel.programming


"Mojo"
<[email protected]>
wrote in message
Hi All
This might be an age old problem, but I just wanted to see if I wasn't
missing something obvious in my code.
Basically I have say 31 people (this figure can range from 10 to 100) who
are put into 8 categories. Once they have been categorised I simply need to
show how many are in each category as a number and as a percentage. Now the
number aspect of display is easy, but I'm coming unstuck with the percetange
aspect.
In essence as I do my % calc for each category sometimes certain values are
1% less or more than they should be and as you add up the displayed figures
the total might come out at 99% or 101%, which looks daft.
An example of a display problem is as follows

1 0 2 9 11 5 0 2 = 30 people
3% 0% 7% 30% 37% 17% 0% 7% = 101% !!!!

My code for each percentage value is simply:
Round((intPeopleCount / intTotalPeopleCount) * 100, 0) & "%"

I really want to make this work, but it seems fraught with issues!!
Any ideas?
Thanks

 

[FSt1]

hi
you are a victim of round off. 1 of your calculation looses .33 and all the
others gain .33 resulting in 1 extra on the total.

regards
FSt1

 

[Arvin Meyer MVP]

You either must show more decimal places or accept the fact that whole
numbers contain portions of others. FWIW, since the computer really doesn't
do decimals (it does binary math) you may still experience some rounding
issues which increase or decrease the percentages. You can make it more
accurate by using the currency datatype which locks the decimals at 4
places. You do not have to format currency data as money, you can format it
any way you wish and use 0 to 4 decimal places.

 

[JoeU2004]

This might be an age old problem
[....]
sometimes certain values are 1% less or more than
they should be and as you add up the displayed
figures the total might come out at 99% or 101%

This is a very common real-world effect of rounding. There are no good ways
to avoid it completely, although there are schemes for ameliorating it.
Professional accounting reports usually have a footnote acknowledging the
fact that rounded values might not add up to the whole.

My code for each percentage value is simply:
Round((intPeopleCount / intTotalPeopleCount) * 100, 0) & "%"

It is unclear if this is VBA code and you are storing this result into
Range.Value, or if this is a formula. If the latter, it would be better to
store a number and use the Percentage format; that is, simply
Round(intPeopleCount/intTotalPeopleCount,2).

I really want to make this work, but it seems fraught with issues!!
Any ideas?

Assuming you store numbers formatted as Percentage, one approach is to put
the following formula into each cell (assuming A2:A9 is intPeopleCount, A10
is intTotalPeopleCount and these formulas are in B2:B9):

=max(0,round(sum($A$2:A2)/$A$10 - sum($B$1:B1),2))

This has the effect of distributing the round-off "error". Note that the
intent is for the relative references A2 and B1 to change in each formula.
And this assumes that B1 is empty, text or zero.

That works with your example, and many others. However, I don't believe it
is a panacea. There may be examples where it still does not work, or it
produces nonsensical results (e.g. non-zero percentage for a category of
zero).

PS: I threw in MAX(0,...) as an insurance policy. I am not sure it is
needed.


----- original message -----

 

[browni]

Hi All

This might be an age old problem, but I just wanted to see if I wasn't
missing something obvious in my code.

Basically I have say 31 people (this figure can range from 10 to 100) who
are put into 8 categories.  Once they have been categorised I simply need to
show how many are in each category as a number and as a percentage.  Now the
number aspect of display is easy, but I'm coming unstuck with the percetange
aspect.

In essence as I do my % calc for each category sometimes certain values are
1% less or more than they should be and as you add up the displayed figures
the total might come out at 99% or 101%, which looks daft.

An example of a display problem is as follows

1       0       2       9        11      5       0      2      = 30 people
3%    0%   7%    30%   37%   17%  0%   7%  =  101%  !!!!

My code for each percentage value is simply:

Round((intPeopleCount / intTotalPeopleCount) * 100, 0) & "%"

I really want to make this work, but it seems fraught with issues!!

Any ideas?

Thanks

Mojo,

If you don't mind where the roundings disappear to here's an approach
that will always give you 100%.

Num: 1 0 2 9 11 5 0 2 30
Cum Num: 1 1 3 12 23 28 28 30
Prev Cum Num: 0 1 1 3 12 23 28 28
Cum %: 3% 3% 10% 40% 77% 93% 93% 100%
Prev Cum %: 0% 3% 3% 10% 40% 77% 93% 93%
% (Cum - Prev Cum): 3% 0% 7% 30% 37% 16% 0% 7% 100%

I hope this gets you started - running totals are the key to this
approach.

Regards,

Ian

 

[Fred]

You might question why you would add up the displayed figures:

- If you want to just show 100%, then just put that in as text without any
calculation

- If you want to verify your underlying data and manipulations, total up the
unrounded numbers.

- There is nothing that says that a total of rounded numbers is supposed to
= 100%, so, so totaling the rounded numbers does not perform any
verification. And, if you do do this, people who have more personality
defects than brains will nitpick your (non-100%) total.

 

[Ian]

You might question why you would add up the displayed figures:  

- If you want to just show 100%, then just put that in as text without any
calculation

- If you want to verify your underlying data and manipulations, total up the
unrounded numbers.  

- There is nothing that says that a total of rounded numbers is supposed to
= 100%, so, so totaling the rounded numbers does not perform any
verification.     And, if you do do this, people who have more personality
defects than brains will nitpick your  (non-100%) total.  

"You might question why you would add up the displayed figures"

Whilst this is a reasonable question in the particular circumstances
of this question, there are, of course, a whole range different
scenarios covering broadly the same issue where it is helpful if the
results displayed add up to exactly what is expected. For example, if
you were dealing with overhead allocations in an accounting context,
you need your journal to balance to the penny. Accounting software
will not allow you to post a journal unless it balances to the penny
(i.e. the debits must be exactly 100% of the credits or vice versa).

There may well be folks looking for answers to some of these different
scenarios that the answers to the original question will be helpful
to.

Regards,

Ian