Re: Sarah Palin - hot or not?

On Tue, 9 Sep 2008 13:56:51 -0700 (PDT), Martin Brown
<|||newspam|||@nezumi.demon.co.uk> wrote:

John Larkin wrote:
On Fri, 5 Sep 2008 08:06:55 -0700 (PDT), mpm <mpmillard@xxxxxxx>
wrote:

On Sep 5, 10:02?am, John Larkin
<jjlar...@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:

If I have serious math to do, I write a program in PBCC, which has
excellent graphics, but is a serious programming language.

True enough, John.
But I do think if kicked the tires on Excel, you would see it does
indeed have tremendous utility.

Take any reasonably complicated modeling task, and I practically
guarantee Excel could get the job done in 10% of the time it would
take to code an alternative approach.

One thing I like (very much) is the ability in Excel to paste
sliderbars on the screen, set their upper & lower bounds, and
increment step size. Then tie this value to a cell - which is later
used in the equations you've set up in Excel.

You can have dozens of such sliderbars set up, each representing some
dependent or independent variable in your model, and move them around
to your heart's content.

Ah, design by fiddling.

You can even graph the results real-time. (Not to mention, you can
also tie your model to real-world, real-time data - such as production
line or testing output.)

Excel also have very powerful Solver applications.
Tell it what variables to consider, and what you want the target
results to be, and it will crunch the numbers for you.

Although Excel can do graphpaper, this is really not pushing the
envelope at all...
Seriously, grab a copy and give it a test drive.
I do think it would be well worth the effort.

-mpm

There are too many learning curves in the world. I'm pretty good with
PADS, LT Spice, assembly, various forms of PowerBasic, Word (ugh!),
fancy text editors, and I have a jillion various utilities I've
accumulated over the years. I don't have time to get good with VHDL or
Excel or Mathcad, so I have minions who do that for me when I need it.

I think you are missing something important here and it is a point
worth making. Because you have to program a spread*** like Excel in
a very different way to normal declarative programming languages an
algorithm done in a spread*** tends not to share the same types of
fence post errors as a classical programming language. Loops are
literally unwound with a row per loop iteration...

Spreadsheets are excellent for tabulating or mocking up test data for
an algorithm or for testing and predicting internal states. Quite a
lot of problems exist where the forward transform is easy and the
backwards one is dificult and what you are trying to program.
Generating test data is something Excel is great at.

I have known scientists (non-programmers) run prototype kit using
combinations of Excel and LabView with surprising ease.

I can get plenty of stuff done by writing small programs, or just
thinking, so I don't really need to use Excel or Mathcad, so I seldom
ask for anyone to do that for me.

I think you would find Excel invaluable as a scratchpad for some
things. It is a lot faster for tabulating and graphing modest amounts
of data. And despite the default marketting oriented graphs (XL2007 XY
graph styles now look like they were drawn by a short sighted 3 year
old with a thick wax crayon) they can be coaxed into a form suitable
for publication in journals.

Sigmaplot or IDL(expensive if you don't qualify for an academic
licence) do a better job if you have a lot of data.

Excel is good when I need a marketing type to, say, graph sales of
some product over time, and do a 2nd order curve-fit on the data, when
said marketing type doesn't actually understand what a 2nd order fit
means. It makes computers useful to people who can't program. *IF* you
sanity check their results.

Excel is better than that - it isn't perfect though.

Don't ask them to do a third order fit in XL2007! The numerical
algorithm in the graph fitting routine has been compromised (now it is
just as bad as LINEST). Previous versions of Excel 2003 and earlier
had a regularised polynomial fit. That is they fit a set of N
orthogonal polynomials to determine the coefficients for x^N very
accurately.

Regards,
Martin Brown


Somehow you are not selling me on the wonder of Excel.

John

.

Quantcast