Re: Continuity correction.


"a.riva@UCL" <anto.ilmago@xxxxxxxxx> wrote in message
news:1193216668.258398.172480@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
On 23 Oct, 19:55, "David A. Heiser" <dah_b...@xxxxxxxxxxxxx> wrote:
"a.riva@UCL" <anto.ilm...@xxxxxxxxx> wrote in message

news:1193155810.098400.152250@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx



On 23 Oct, 17:03, Ken Butler <but...@xxxxxxxxxxxxxxxx> wrote:
On Tue, 23 Oct 2007 15:44:31 -0000, "a.riva@UCL"

<anto.ilm...@xxxxxxxxx> wrote:
So, this means that if I want to calculate the probability associated
to X>=n (with X positive this corresponds to the right tail of the
distribution) I should convert everything to the calculation of
P(Z>=n-0.5) so that I include n. And using NORMSDIST(Abs(z)), the
cumulative probability left to n-0.5 is NORMSDIST(Abs(z(n-0.5))), and
the probability associated with the right tail is 1-
NORMSDIST(Abs(z(n-0.5))). Is this right?

I don't claim to be an expert in NORMSDIST, but it looks right to me.
--
Ken Butler, Lecturer (Statistics)
University of Toronto at Scarborough
butler (at) utsc.utoronto.cahttp://www.utsc.utoronto.ca/~butler

Great!!!

Thanks a lot!!! Unfortunately Excel is really limited for statistical
analyses, but it's generally the software that we use to create all
our databases, and sometimes it's really quicker to try to implement
it instead of moving all the data in other softwares...

Cheers,

Antonio.

++++++++++++++++++++++++++++++++++++++++
Actually for Excel 2003 and 2007, NORMSDIST is accurate from z values
of -37.5 upward. The lower limit is from the lower limit of the IEEE-754
double. There is a region of around z from -3 to -8 where the accuracy
drops
to 6 significant decimal digits. Outside of this the accuracy is in the
region of 14 decimal digits.

NORMSDIST has been pretty well tested. See my URL on this (search for
"Excel
Faults")

What is your rational for saying that Excel is limited here?

David Heiser- Hide quoted text -

- Show quoted text -

Hi David.

Sorry for my comment about Excel. I didn't explain accurately what I
mean.
I was not actually expressing doubts about the accuracy of the
functions that Excel already contains. I have to say that I find them
really useful and definitely precise and accurate. What I exactly
meant is that some statistical tests or functions or graphs are not
included in Excel. For example, non parametric tests, GLMs, boxplots,
ranking function which takes ties into consideration. If one wants to
perform them, they have to be "built" combining the existing
functions. This is absolutely fine, but not always straightforward :-)
It's true that Excel wasn't born as a statistical analysis software,
but it's more and more used also for this purpose.
++++++++++++++++++++++++++++++++++++++++++++++++
Yes you are right here. Excel is very limited on statistical
functions/routines.

From about 1998 on, it was very popular to write new introductory stat
textbooks that included use of Minitab, Excel, et. al, to solve problems.
Publishers loved this, because they could include a cdrom with all the
problems sets, and their own add-ins to do things that the software could
not do. They now had a reason to increase textbook costs, and create a
market to sell their add-in software. All the hidden stat-teacher-authors
now had a reason to revise/rewrite their texts for new editions, making the
used texts no longer useable. Great marketing ploy.

The text books were a hidden means to sell software. This is still
continuing today.

MIcrosoft viewed Excel as a business tool. The stat functions and routines
were just incidental add-ins that were in the original Graymatter software
that Microsoft bought, since the focus was on the financial functions it
contained. The stat is all 1960's stuff and has never been updated as to
context.

The wide use of Microsoft Office, essentially has created an international
body of users, that rely on Excel and the functions to solve their problems.

There are hidden issues regarding "competition", "copy-rights", "algorithm
patents" and "market dominance" (especially in Europe), that make the issue
of expanding Excel's stat capabilities (in competition to all the other
commercial stat software and Excell add-ns) an uncertain area to proceed in.
Microsoft would rather list available commercial add-ins. They see no
marketing advantage to add to Excel's stat capabilities.

So there you have it, a 1960's view of statistics, locked into all future
time.

If business pushed Microsoft to add to the stat capabilites, Microsoft
probably would. However, that is not what business' want. Just spend some
time on the Excel related lists.

Like everything else in our world, it is only a matter of dollars.

David Heiser
++++++++++++++++++++++++++++++++++++++++++++++++=





Just this.

By the way, could you explain to me what is the logic behind the fact
that NORMSDIST returns the cumulative probability "left-to-right"
whilst TDIST returns the cumulative probability "right-to-left"?

Thanks, and I hope I explained myself better now :-)

Antonio.
++++++++++++++++++++++++++++++++++++++++++
No I can't. It all comes from historic usage, of negative being on the left
and positive being on the right. The normal is viewed as a stand-alone
concept. The t distribution only as a test, not a visual concept of the
distribution. The F distribution shares this too. The chi-square is like the
normal, a stand-alone distribution.

David Heiser
+++++++++++++++++++++++++++++++++++++++++


.

Relevant Pages

  • Re: Find a distribution!
    ... >> I am working on a project at a company in Sweden where the amount of ... >> bacterias in food are simulated in Excel. ... > The exponential and the gamma distribution I think can give you quit ... If you're looking at pathogens in food it may be that you ...
    (sci.stat.math)
  • Re: Autocreating probability distribution
    ... this functionality is not built into Excel. ... In the case of the Normal distribution, the NORMDIST function takes its mean ... and standard deviation as arguments. ... Excel probability functions are not consistent as to whether they calculate ...
    (microsoft.public.excel)
  • Re: How to generate random X given only min, max, mean?
    ... A uniform distribution from 74 to 76 would meet ... Not that any of the Excel random number generators ... McCullough's tests on the RN generators, ...
    (sci.stat.edu)
  • Re: Random Number Generation Dialog Box
    ... >> Am using Excel 2001 and want to generate some random numbers from a normal ... >> distribution. ... A help search tells me about what looks like a wonder dialog ... > Be aware, however, that the ATP's pseudorandom number generator is not ...
    (microsoft.public.mac.office.excel)
  • Re: how to find z scores in statistics
    ... Excel has a number of statistical functions built in. ... X is the value for which you want the distribution. ... Standard_dev is the standard deviation of the distribution. ...
    (microsoft.public.excel.worksheet.functions)