| |
 |
|
|
Science Forum Index » Statistics - Education Forum » Studentized range (q)
Page 1 of 1
|
| Author |
Message |
| Stan Brown |
 Posted: Sat Mar 29, 2008 4:00 pm |
|
|
|
Guest
|
I'd like to program Tukey's HSD test into my TI-84, but to do that I
need to come up with the critical values of the q distribution. In
googling, I've been unable to find any description of the q
distribution, though there are lots of tables of critical values.
Can the q distribution be derived in terms of the normal, Student's
t, Chi-squared, and F distributions, which are available on my
calculator?
Thanks in advance!
--
Stan Brown, Oak Road Systems, Tompkins County, New York, USA
http://OakRoadSystems.com
Shikata ga nai... |
|
|
| Back to top |
|
| Richard Ulrich |
| Posted: Sat Mar 29, 2008 8:18 pm |
|
|
|
Guest
|
On Sat, 29 Mar 2008 17:00:48 -0400, Stan Brown
<the_stan_brown@fastmail.fm> wrote:
Quote:
I'd like to program Tukey's HSD test into my TI-84, but to do that I
need to come up with the critical values of the q distribution. In
googling, I've been unable to find any description of the q
distribution, though there are lots of tables of critical values.
Can the q distribution be derived in terms of the normal, Student's
t, Chi-squared, and F distributions, which are available on my
calculator?
No, I don't think it is that simple.
Here is a curious anecdote about Tukey's HSD test.
In 1981 or so, I wrote a program to provide the test, making
use of the "IMSL subroutine package", which was a popular
product that was available for scientific computing on
all mainframes. IMSL accepted a vector of means, etc.,
and returned information about which sets were not-different.
They cited a particular textbook (perhaps Kush? Kirk?), and
the implementation could be validated using an example
of data from the text.
A decade later, without making mention of making a change,
they re-wrote the routine. However, new documentation
pointed to a different textbook, and a different numeric
example.
What is curious is that the means that were "different"
according to the first text and implementation were no longer
different; and, similarly in the other direction, the old
program did not find the "differences" that were shown by
the new one.
This roused my curiosity, but I never did find basic
documentation to tell me what was going on. I would be
pleased to learn, even now.
--
Rich Ulrich
http://www.pitt.edu/~wpilib/index.html |
|
|
| Back to top |
|
| Ray Koopman |
| Posted: Sun Mar 30, 2008 2:10 pm |
|
|
|
Guest
|
On Mar 29, 2:00 pm, Stan Brown <the_stan_br...@fastmail.fm> wrote:
Quote: I'd like to program Tukey's HSD test into my TI-84, but to do that I
need to come up with the critical values of the q distribution. In
googling, I've been unable to find any description of the q
distribution, though there are lots of tables of critical values.
Can the q distribution be derived in terms of the normal, Student's
t, Chi-squared, and F distributions, which are available on my
calculator?
Thanks in advance!
--
Stan Brown, Oak Road Systems, Tompkins County, New York, USA
http://OakRoadSystems.com
Shikata ga nai...
http://lib.stat.cmu.edu/apstat/190
Algorithm AS 190 Appl. Statist. (1983) Vol.32, No. 2
Incorporates corrections from Appl. Statist. (1985) Vol.34 (1)
Evaluates the probability from 0 to q for a studentized
range having v degrees of freedom and r samples.
Algorithm AS 190.1 Appl. Statist. (1983) Vol.32, No. 2
Approximates the quantile p for a studentized range
distribution having v degrees of freedom and r samples
for probability p, p.ge.0.90 .and. p.le.0.99
Algorithm AS 190.2 Appl. Statist. (1983) Vol.32, No.2
Calculates an initial quantile p for a studentized range
distribution having v degrees of freedom and r samples
for probability p, p.gt.0.80 .and. p.lt.0.995 |
|
|
| Back to top |
|
| Stan Brown |
| Posted: Sun Mar 30, 2008 4:10 pm |
|
|
|
Guest
|
Sat, 29 Mar 2008 21:18:11 -0400 from Richard Ulrich
<Rich.Ulrich@comcast.net>:
Quote: On Sat, 29 Mar 2008 17:00:48 -0400, Stan Brown
the_stan_brown@fastmail.fm> wrote:
Can the q distribution be derived in terms of the normal, Student's
t, Chi-squared, and F distributions, which are available on my
calculator?
No, I don't think it is that simple.
Oh darn! Well, thanks anyway.
[big snip]
Quote: This roused my curiosity, but I never did find basic
documentation to tell me what was going on. I would be
pleased to learn, even now.
Me too. Anyone have an algorithm? Surprisingly, the q distribution
isn't even mentioned at Mathworld, and all the references I turned up
by googling just referred to a table lookup. As Robert Heinlein said
of cube roots, "Those tables weren't brought down from heaven by an
archangel."
--
Stan Brown, Oak Road Systems, Tompkins County, New York, USA
http://OakRoadSystems.com
Shikata ga nai... |
|
|
| Back to top |
|
| Stan Brown |
| Posted: Mon Mar 31, 2008 4:07 am |
|
|
|
Guest
|
Sun, 30 Mar 2008 17:10:00 -0700 (PDT) from Ray Koopman
<koopman@sfu.ca>:
Quote: On Mar 29, 2:00 pm, Stan Brown <the_stan_br...@fastmail.fm> wrote:
Can the q distribution be derived in terms of the normal, Student's
t, Chi-squared, and F distributions, which are available on my
calculator?
http://lib.stat.cmu.edu/apstat/190
Algorithm AS 190 Appl. Statist. (1983) Vol.32, No. 2
Incorporates corrections from Appl. Statist. (1985) Vol.34 (1)
Evaluates the probability from 0 to q for a studentized
range having v degrees of freedom and r samples.
Algorithm AS 190.1 Appl. Statist. (1983) Vol.32, No. 2
Approximates the quantile p for a studentized range
distribution having v degrees of freedom and r samples
for probability p, p.ge.0.90 .and. p.le.0.99
Algorithm AS 190.2 Appl. Statist. (1983) Vol.32, No.2
Calculates an initial quantile p for a studentized range
distribution having v degrees of freedom and r samples
for probability p, p.gt.0.80 .and. p.lt.0.995
Thanks, Ray! At first glance it looks like it may be too complicated
for my calculator, but I'm going to look more closely in hopes of
finding blocks of code that can be replaced by native TI functions.
Thanks also in general for pointing me toward this library of
algorithms!
--
Stan Brown, Oak Road Systems, Tompkins County, New York, USA
http://OakRoadSystems.com
Shikata ga nai... |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Thu Jan 08, 2009 12:56 pm
|
|