xl 2003 - Analysis ToolPak question

[Dana DeLouis]

I would like to ask for an opinion on the "Random Number Generation"
function of the Analysis ToolPak in Excel 2003. I have all the latest
updates for the program.
If I select "Random Number Generation", and select the "Patterned"
distribution, from 1 to 12 in steps of 3, I get...

1, 4, 7, 10, 12.

I wasn't expecting the ending 12. I was thinking it should be just 1, 4, 7,
10. The next number should be 13, but since 13 is outside the limits, it
would not be included.

Do any of you "Stat" guys have any thoughts on this? In statistics, is it
normal to include the ending number range even though it doesn't fit the
"pattern"? I'm just curious, that's all.

VBA would look like this:

Application.Run "ATPVBAEN.XLA!Random", ActiveSheet.Range("A1"), , , 6,
, 1, 12, 3, 1, 1

The "6" is the 6th item in the list ("Pattern") running from 1 to 12 in
steps of 3.
I get
1, 4, 7, 10, 12.

Thanks for any insight. Again, I am just curious. Didn't see anything in
the kb articles to shed light.

Dana DeLouis

 

[hgrove]

Dana DeLouis wroet...

I would like to ask for an opinion on the "Random Number Generation function
of the Analysis ToolPak in Excel 2003. I have all the latest update for the
program. If I select "Random Number Generation", and select th "Patterned"
distribution, from 1 to 12 in steps of 3, I get...

1, 4, 7, 10, 12.

I wasn't expecting the ending 12. I was thinking it should be just 1 4, 7, 10.
The next number should be 13, but since 13 is outside the limits, i would not be
included.

...

It's fubar in older versions too. Looks like it's written to includ
the end points no matter what the step value may be. After all, 1 to 1
in steps of 3 could just as easily mean 3,6,9,12 rather than 1,4,7,10.

Where's Dave Braden when we need him. He's written a lot about th
failings of Excel's discrete distribution functions. This would be jus
another example.

Is this 'wrong'? In the same sense that =-5^2 gives +25 rather than -2
in Excel, which is at odds with the standard mathematical precedenc
convention, is wrong, this is wrong. To the extent that Microsoft ca
(and generally does) do precisely what it damn well pleases, this i
consistent with program specs (perhaps in the sense that an
proposition P is consistent with an empty set of axioms). Best t
consider it just another quirk that you'll have to live with, remembe
and work around until Microsoft fixes this (unlikely to happen in thi
lifetime or the next)

 

[Dana DeLouis]

Thanks for the feedback Harlan. I didn't know what to make of it. You're
right though. Under the heading of "Random Number Generation" I thought it
was suppose to randomly return 1,4,7,10, or 2,5,8,11, or as you said
3,6,9,12.
Oh well. Not a big deal. I was just curious.
If I put in 4 for the "Number of Random Numbers:", I still get back 5
numbers ... 1,4,7,10,12.
Hmmm.

Thanks for the feedback. :>)
Dana DeLouis

 

[Jerry W. Lewis]

There seems to be a couple of different issues here

Random Number Generation:
My understanding is that Excel contains three distinct random number
generators (RNGs): worksheet, ATP, and VBA. It is my understanding that
only the worksheet RNG was upgraded in 2003, with the result that
neither the ATP nor the VBA RNG is suitable for serious work even in
Excel 2003.

Patterned Distribution:
I have nothing to add to Harlan's discussion of the interpretation of
the discrete distribution from 1 to 12 in steps of 3, but notice that
this particular functionality has nothing to do with random numbers,
despite the menu title that selects to it.

Jerry