How can one get in touch with the designers of Excel?

[Omnibus]

Hello Everyone,

This is my first time posting a message in this forum and I hope this here is the right section to post it: Does anyone know who the concrete programmers who designed Excel (say, Excel 2010) within Microsoft are and how one can get in touch with them? Thanks in advance.

 

[shg]

Numerical analysis is the topic that deals with errors accumulated from approximate solutions (e.g., using finite-precision floats). There is a world full of literature available.

 

[Omnibus]

Omnibus

I don't see how the people that created Excel would be more knowledgable about this than a mathematician(s).

It's a simple fact that computers cannot represent real numbers precisely using floating-point methods and can't really 'do' precise arithmetic with them.

I understand that but nevertheless it puzzles me that something that comes out to be practically the same when calculated in five different ways becomes something else when the sixth way is applied (sixth is expected to yield precisely the same result as the other five). Mind you, all six calculations give the same result after a slight change in the formula.

 

[Norie]

What has that got to do with Excel?

I don't think you'll find Excel using any significantly different 'alogorithims' than any other similar package, or even 'advanced' mathematical tools.

 

[Omnibus]

What has that got to do with Excel?

I don't think you'll find Excel using any significantly different 'alogorithims' than any other similar package, or even 'advanced' mathematical tools.

I haven't tried other spreadsheets but Mathematica, for instance, gives always the same result for all six calculations. It probably has something to do with the algorithm whereby any sum is treated as a Riemann sum which may not always be the case. That's just a guess. Excel, at least, is somewhat more transparent as to the arithmetics.

 

[shg]

Is there a particular Excel function involved in this?

 

[Omnibus]

Is there a particular Excel function involved in this?

Probably it would help if I give you the exact formulae I’m using in the spreadsheet:

Column A is =A3+0.00000000125 where A3 = 0. Values stretch down to cell 1003.

Column B of the spreadsheet =(1.17/SQRT(9.9244^2+(1/(2*PI()*800000*0.000000000115))^2))*SIN((2*PI()*800000*A3)+
ATAN(1/(9.9244*2*PI()*800000*0.000000000115)))

ColumnC =-3.356+(1.17*SIN((2*PI()*800000*A3)))

An example of what each cell in column D calculates is given by cell D3 =B3*C3, Cell D4 will be =B4*C4 and so on.

Same for cell E3 where the calculation is =B3*B3*9.9244. Thus, cell E4 will be = B4*B4*9.9244 and so on.

The cells in F and G columns calculate, respectively, =I2+(ABS((A4)-(A3))*(((D3)+(D2))/2)) and
=J2+(ABS((A4)-(A3))*(((G3)+(G2))/2)) all pasted down to 1003rd row, as above.

The H4 and I4 cells calculate, respectively, =(F1003-F3)/(A1003-A3) and =(G1003-G3)/(A1003-A3)

Cells H5 and I5 calculate, respectively, =AVERAGE(D3:D1003) and =AVERAGE(E3:E1003)

At the end the ratios = I4/H4 and = I5/H5 are calculated in cells J4 and J5, respectively.



Now, notice, the result obtained in cells H4, I4 and I5 practically coincide:

H4 -- 0.000002269692833487160000000000​

I4 -- 0.000002269692833467220000000000​

I5 -- 0.000002271960109632750000000000​

while the result in cell H5 is dramatically different

H5 -- 0.000000000030645126327480400000​

Replace now the formula =-3.356+(1.17*SIN((2*PI()*800000*A3))) in column C by =(1.17*SIN((2*PI()*800000*A3))) and you will see that all cells (H4, I4, H5 and I5) show practically the same result.

 

[shg]

I don't know what your example illustrates (presumably Riemann integration), but there may be better ways (i.e., ways that converge faster) to do numerical integration.

The cells in F and G columns calculate, respectively,

=I2+(ABS((A4)-(A3))*(((D3)+(D2))/2)) and

=J2+(ABS((A4)-(A3))*(((G3)+(G2))/2))

all pasted down to 1003rd row, as above.

Seems that the references to I2 and J2 must be wrong. There's nothing in those cells.

Replace now the formula =-3.356+(1.17*SIN((2*PI()*800000*A3))) in column C by =(1.17*SIN((2*PI()*800000*A3)))

The second formula is intrinsically better; in the first, you're adding numbers different by orders of magnitude from each other, made to order to lose significant digits.

 

[Omnibus]

That's a typo. The formuli should read as follows:

The cells in F and G columns calculate, respectively,

=F2+(ABS((A4)-(A3))*(((D3)+(D2))/2)) and

=G2+(ABS((A4)-(A3))*(((G3)+(G2))/2))

all pasted down to 1003rd row, as above.

 

[Omnibus]

Quote:
Replace now the formula =-3.356+(1.17*SIN((2*PI()*800000*A3))) in column C by =(1.17*SIN((2*PI()*800000*A3)))
. . .
The second formula is intrinsically better; in the first, you're adding numbers different by orders of magnitude from each other, made to order to lose significant digits.

Why should adding a small number to a large number make a difference? I'm away from my laptop right now and the iPad I'm using isn't the best for this purpose. As soon as I get back I'll try to play little bit with this column to understand what you really mean. It escapes me at this point.

 

[Macropod]

Why should adding a small number to a large number make a difference? I'm away from my laptop right now and the iPad I'm using isn't the best for this purpose. As soon as I get back I'll try to play little bit with this column to understand what you really mean. It escapes me at this point.

Because one of the limitations in Excel is that it can only store numbers 15-digits long.

As for getting in touch with the MS programmers, I doubt MS would let you do that. The nearest you'll get is MS Tech Support.

You might find the observations in following links eye-opening.
www.forecastingprinciples.com/paperpdf/McCullough.pdf
http://www.practicalstats.com/xlsstats/excelstats.html