Newsgroup sci.stat.math 12221

Directory

On Fri, 13 Dec 1996 00:07:22 -0500, Ya-Fen Lo 
wrote:
>Hi,
>
>This is a beginners' SAS question.
>I am a social scientist trying to 
>finish my final project in a research class.
>
>Is it possible to perform tests of simple effects
>(as defined in APPLIED STATISTICS by HINKEL/WIERSMA/JURS)
>in SAS ? I am using the following setup
>
>PROC ANOVA DATA=PROJECT;
>     CLASSES A B;
>     MODEL S=A B A*B;
>     MEANS A B A*B
>     MEANS A B A*B/TUKEY BON;
>     FORMAT A AA. B BB.;
>     TITLE 'THE TWO-WAY FIXED-MODEL ANOVA';
>
>The second means statement doesn't perform the simple
>effects as I would have expected.
>
You can use the TEST statement in proc GLM. May be also in proc ANOVA. Do you
simply want to know if an effect is significant? if so, just check the ANOVA
table.
I suppose this is just a typing error but CLASSES should be written CLASS.
R

I know it's a long list, but could someone recommend a book(s) that 
would contain good reference material on Cluster Analysis, Market 
Segmentation, and Neural Networks and preferrably on how to use these 
techniques to do modeling.
Thanks for your E-mail in advance.

This would seem to beg the question, because it's far from clear 
how "working from data" would allow estimation of p(AorBorC) any more 
directly than of p(ABC) itself.  (One would usually use your formula to 
estimate p(AorBorC) from p(ABC), etc. , not vice versa.)
Robert E Sawyer 
soen@pacbell.net
____________________
BARNEY KREBS  wrote in article <32b5f7a4.12253664@news.in.net>...
|Theodore Sternberg  wrote:

| >Suppose we know the three marginal probabilities, as well as all
| >the two-way joint probabilities p(A,B), p(B,C) and p(A,C). 
| >But we don't know the three-way joint probabilities.  What is the most
| >"natural" guess at the three-way probabilities?  

|
| Dumb Suggestion:
| 
| If you're working from data then you should be able to estimate  
| P(A or B or C).  Then, using the inclusion exclusion principle,
| 
| P(ABC) = P(A or B or C) - P(A) - P(B) - P(C) + P(AB) + P(AC) + P(BC).

Dear All,
I would appreciate some input to the following. Thanks.
I have a set of data points (Xi,Yi,Zdata) and want to compare
them to the predictions of two models (Xi,Yi,Zone) and (Xi,Yi,Ztwo).
For each Xi,Yi, we can form an ordered sample (Zdata,Zone) and
(Zdata,Ztwo) where the Xi,Yi are matched. The ordered samples
can be drawn as a scatter diagram.
I then calculate a measure of the agreement between the data and
each of the models, by finding the mean of the squared errors of
prediction (Sigma(from i to n)(Yi - Yi')^2)/n where Yi' is the model
prediction, Yi is the observed data point, i is each combination of
Xi,Yi and n is the number of paired observations (data,model prediction).
(I don't want to measure the correlation of the scatter diagram points
as this just gives me a measure of the linearity - I want to find
the agreement between the paired samples and some given line defined
by model one and model two).
Anyway after doing this I get a mean of the squared errors for
Zdata to Zone (call it A), and another mean of the squared errors for 
Zdata to Ztwo (call it B). B is around seven times larger than A.
The question is (if you've had the patience to get this far), is there
a way of measuring the statistical significance of the fact that B is
much larger than A? Or am I hoping for too much and does it all boil down
to the fact that a correlation-variant measure that I've done won't
support assertions of this kind?
Thanks for your time.
Best wishes - David.
dsmith@psy.ucsd.edu

blosskf@apci.com (Karl F. Bloss) writes:
}
} * Numerical Recipes in C/FORTRAN
checker@netcom.com (Chris Hecker) writes:
>
>Anyone thinking of using the algorithms NR should look at this page:
>
>http://math.jpl.nasa.gov/nr/
>
>The page starts with, "We have found Numerical Recipes to be generally
>unreliable," and then goes on to show why.
 Then read the NR page that answers the criticism, but while at NR, 
 read their *licensing* rules.  All of their code is copyrighted 
 and even if you buy the PC code you cannot migrate it to a UNIX 
 box without paying the fee.  Legally, that is.  Special clearance 
 is required to include their code in a copyrighted thesis.  Etc. 
 Then visit Netlib and see what is available for free. 
 If you want to learn algorithms, there are better books as you see 
 mentioned on that web page.  If you want code, there is lots out 
 there of very high quality.  Compare NR with the alternatives in a 
 good bookstore to be sure it meets your needs. 
-- 
 James A. Carr        |  "The half of knowledge is knowing
    http://www.scri.fsu.edu/~jac/       |  where to find knowledge" - Anon. 
 Supercomputer Computations Res. Inst.  |  Motto over the entrance to Dodd 
 Florida State, Tallahassee FL 32306    |  Hall, former library at FSCW.