Statistical Bug?

[Trent Moseley]

Is there a bug in the standard deviation statistical function of some HP machines but not in others? Of the HPs I have with the stat function the 12C,15C, and the 42S seem to have the bug. The 25C, 67, and the 32SII do not. Here is the problem:

With the summation key enter in several data and find the standard deviation. Next subtract a datum. Finding the SD now gives an error message.

c

cC

[R Lion (Spain)]

I don't know if I understand you.
I usually get mean, then sigma+ and get SD. Then mean again, sigma-. If I get SD now, I don't get any error message.

Could you give any example?

Raul

[Trent Moseley]

For example:

clear sigma+

9 sigma+ see 1

15 sigma+ see 2

11 sigma+ see 3

take std dev see 3.06

11 sigma- see 2

take std dev see Error 2 on the 12C, 15C
Stat Math Error on the 42S

You do not get an error message on the 25C, 67, or 32SII but see the answer: 4.24

[R Lion (Spain)]

You can avoid the problem introducing 0 as y data:

0 enter 9 sigma+ see 1

0 enter 15 sigma+ see 2

0 enter 11 sigma+ see 3

etc... and you'll get the expected answer

Raul

[R Lion (Spain)]

When you get one of those strange errors, STOre a 0 in regs 5, 6 and 7 (sumy, sumy^2 and sumxy). Then, s will work

Raul

[Trent Moseley]

Raul Lion,

I don't bother. I use one of the machines that don't have the problem.

tm

[R Lion (Spain)]

Oh well... then send me your 15c and I'll send you a new 32sII ;-)

Raul

[Trent Moseley]

Raul,

I love 'em all! The 16C didn't even get into the discussion! Now if only I could get my hands on a 34C!

Trent

[Les Bell]

It happens as you describe on the 41CX, as well, but the reason is simple - it's the data in the Y register. When you get the SDEV result in X, then simply key in 11 and press sigma-, it will have 3.06 in the Y register, which it subtracts from the cumulative stat registers (which will have zero for both sigma Y and sigma(Y^2), from memory).

If you key in a zero before the 11 to be sigma-minussed, or clear the stack, it will work just fine.

Best,

--- Les [http://www.lesbell.com.au]

[Gene]

Think you'll need to give more specifics.

Which mode is the 42s in? LinSUM or ALLSUM?

Did you do a CLEARSUM before entering data?

Which data points did you enter?

Which did you remove?

Maybe these will help us help you.

[Trent Moseley]

Gene,

See my reply to R. Lion

tm

[Karl Schneider]

Trent --

There is no bug. Les Bell's reply below hit the nail on the head -- not clearing the previous SDEV calculation caused it to be entered as a y-datum deleted from summation for two-variable analysis upon doing the Sigma-minus. SDEV of the y-data was undefined (sqrt of negative number).

The 32Sii did not yield an error because it has separate keystrokes for x-SDEV and y-SDEV, population or sample. The 12C, 15C, and 42S will calculate both sample SDEV's with the same keystroke. The older 25C and 67 might not have two-variable stat analysis built in (I have neither unit).

-- Karl