Subject: Characterization of multivariate dsn
From: Theodore Sternberg
Date: 19 Dec 1996 18:30:52 GMT
Do there exist multivariate distributions that, like the normal, are fully
characterised by their first two moments, but unlike the normal are easily
evaluated (even in high dimensions)? My dream distribution would look as
"normal" as possible, but be easy to evaluate.
Would I need to look for an exponential-family distribution whose
sufficient statistics involve only sums, sums of squares and cross-sums?
I suppose an infinite number of such beasts exist, and so the question is
how do I cook up one that has an easy-to-evaluate cumulative distribution?
Am I on the right track here?
Ted Sternberg
San Jose, California USA
Subject: eigenvectors of rotated solution
From: "Ralf Schulze"
Date: 19 Dec 1996 19:03:07 GMT
Hello,
I'm trying to find the eigenvectors corresponding to the factors
of a (varimax-)rotated principal components solution. I have three factors,
the transformation matrix, the unrotated solution of the factor pattern and
the "rotated eigenvalues" from the rotated factor pattern. I tried to
compute
the eigenvectors for the rotated solution via:
F*D**-.5 = A
where A is the matrix of the eigenvectors, F is the rotated factor pattern
and D**-.5 is the inverse of a diagonal matrix with the square roots of the
eigenvalues in the principal diagonal.
This works fine for the unrotated solution (for which I have the
eigenvectors, of course) but not for the rotated solution cause the
resulting
eigenvectors are not mutually orthogonal (I think they have to be
orthogonal,
but A'A is not diagonal).
This is probably a simple question for you but I really can't figure
out why the resulting eigenvectors are not orthogonal and therefore
this simple formula doesn't work.
Thanks in advance for any helpful comments.
--
Ralf Schulze
LS II Psychologie
Universität Mannheim
EMail: schulze@tnt.psychologie.uni-mannheim.de
Subject: Re: Simulation - Monte Carlo?
From: radford@cs.toronto.edu (Radford Neal)
Date: 19 Dec 96 19:47:05 GMT
In article <32B85621.13D36AFF@ulst.ac.uk>,
JG.Campbell wrote:
>I used to use the term 'Monte Carlo simulation' for simulation
>procedures like that described below although I'm fairly sure now that
>this is an abuse of the term...
>Simulation: we know the probability density/distribution of q, so,
>using appropriate random number generators, we generate values qi, and
>compute yi = f(x;qi).
>
>Using a large number of iterations, we estimate the distribution of y
>-- or some statistics. We used to call this a Monte Carlo 'loop'.
>
>Hence questions: (a) is this 'Monte Carlo simulation' in any reasonable
>interpretation of the term? (b) If not, is there another appropriate
>term?
It is quite correct to describe this as a "Monte Carlo simulation".
There term is rather general. This might also be the best way to
do what you're doing, especially if q is high dimensional, and the
function f is complicated.
----------------------------------------------------------------------------
Radford M. Neal radford@cs.utoronto.ca
Dept. of Statistics and Dept. of Computer Science radford@utstat.utoronto.ca
University of Toronto http://www.cs.utoronto.ca/~radford
----------------------------------------------------------------------------
Subject: Job Announcements - Jr. and Sr. Positions
From: chappell@becrux.biostat.wisc.edu (Rick Chappell)
Date: 19 Dec 1996 21:20:28 GMT
Job Openings Announcement - The University of Wisconsin Medical School
Department of Biostatistics
The U.W. Dept. of Biostatistics is seeking candidates for two tenure track
biostatistician positions - one Assoc. or Full professor and one Asst.
Professor. Candidates should have PhD in biostatistics or statistics and track
record in teaching, statistical & collaborative research. Responsibilities
include teaching, research, & appropriate univ. & professional service.
Applicants: send resume and 3 letters of reference to David DeMets, Chair,
Dept. of Biostatistics, University of Wisconsin-Madison, 600 Highland Avenue,
Room K6/446, Madison, WI 53792-4676. Applications will be accepted until the
position is filled. AA/EOE. Women & minorities are encouraged to apply.
Unless confidentiality is requested in writing, information regarding
applicants must be released upon request. Finalists cannot be guaranteed
confidentiality.
Subject: Need code for an ARIMA (1,0,1) Model
From: catullus@laraby.tiac.net (Robert Kelley)
Date: 19 Dec 96 23:27:11 GMT
Hello:
I am searching for source code that calculates an ARIMA(1,0,1) model.
I would prefer if it is was in some flavor of Basic, but am willing to
look at any source code that can do it.
I would even use a stat-library; but would prefer not to use an
executable. I need to to stick an ARIMA (1,0,1) Model into pre-existing
source code.
Any leads on this would be *greatly* appreciated.
catullus@laraby.tiac.net
--
_______________________________________________________________________
Robert W. Kelley (http://www.tiac.net/users/rkelley/) "odi et amo..."
Nothing makes one so vain as being told that one is a sinner.
Conscience makes egotists of us all.
Subject: Modern Regression and Classification course - Hawaii
From: Trevor Hastie
Date: 19 Dec 1996 16:42:11 -0800
************* 1997 Course Announcement *********
MODERN REGRESSION AND CLASSIFICATION
Waikiki, Hawaii: February 17-18, 1997
*************************************************
A two-day course on widely applicable statistical methods for
modeling and prediction, featuring
Professor Trevor Hastie and Professor Robert Tibshirani
Stanford University University of Toronto
This course was offered and enthusiastically attended at five
different locations in the USA in 1996.
This two day course covers modern tools for statistical prediction and
classification. We start from square one, with a review of linear
techniques for regression and classification, and then take attendees
through a tour of:
o Flexible regression techniques
o Classification and regression trees
o Neural networks
o Projection pursuit regression
o Nearest Neighbor methods
o Learning vector quantization
o Wavelets
o Bootstrap and cross-validation
We will also illustrate software tools for implementing the methods.
Our objective is to provide attendees with the background and
knowledge necessary to apply these modern tools to solve their own
real-world problems. The course is geared for:
o Statisticians
o Financial analysts
o Industrial managers
o Medical and Quantitative researchers
o Scientists
o others interested in prediction and classification
Attendees should have an undergraduate degree in a quantitative
field, or have knowledge and experience working in such a field.
PRICE: $750 per attendee if received by January 15, 1997. Full time
registered students receive a 40% discount. Attendance is limited to
the first 60 applicants, so sign up soon! These courses fill up
quickly.
TO REGISTER: Fill in and return the form appended.
For more details on the course and the instructors:
o point your web browser to:
http://stat.stanford.edu/~trevor/mrc.html
OR send a request by
o FAX to Prof. T. Hastie at (415) 326-0854, OR
o email to trevor@stat.stanford.edu
<----------------------------- Cut Here ------------------------------->
Please print, and fill in the hard copy to return by mail or FAX
REGISTRATION FORM
Modern Regression and Classification
Monday, February 17 and Tuesday, February 18, 1997.
Hilton Hawaiian Village, Waikiki Beach, Honolulu, Hawaii.
Name ___________________________________________________
Last First Middle
Firm or Institution ______________________________________
Standard Registration ____ Student Registration ____
Mailing Address (for receipt) _________________________
__________________________________________________________
__________________________________________________________
__________________________________________________________
Country Phone FAX
__________________________________________________________
email address
__________________________________________ _______________
Credit card # (if payment by credit card) Expiration Date
(Lunch preference - tick as appropriate):
___ Vegetarian ___ Non-Vegetarian
Fee payment can be made by MONEY ORDER , PERSONAL CHECK, or CREDIT CARD
(Mastercard or Visa.) For checks and money orders: all amounts are given in
US dollar figures. Make fee payable to Prof. T. Hastie. Mail it, together
with this completed Registration Form to:
Prof. T. Hastie
538 Campus Drive
Stanford
CA 94305
USA
For payment by credit card, include credit card details above, and mail to
above address, or else FAX form to 415-326-0854
For further information, contact:
Trevor Hastie
Stanford University
Tel. or FAX: 415-326-0854
e-mail: trevor@stat.stanford.edu.
http://stat.stanford.edu/~trevor/mrc.html
REGISTRATION FEE
Standard Registration: U.S. $750 ($950 after Jan 15, 1997)
Student Registration: U.S. $450 ($530 after Jan 15, 1997)
Student registrations - include copy of student ID.
- Cancellation policy: No fee if cancellation before Jan 15, 1997.
- Cancellation fee after January 15 but before Feb 12, 1997: $100.
- Refund at discretion of organizers if cancellation after Feb 12, 1997.
- Registration fee includes course materials, coffee breaks, and lunches
- On-site Registration is possible if course is not fully booked, at late
fee.
Subject: Re: multicollinearity
From: jim bouldin
Date: Thu, 19 Dec 1996 23:10:42 -0800
Jim Bouldin wrote:
> > So, question one. Would a solution be an analysis of covariance by
> > turning one of the two continuous ind variables into a categorical one
> > and using it as a covariate? Within each category the correlation
> > between the two ind vars should be greatly reduced, right? So I could
> > produce an estimate of the independent effects of each ind variable on
> > the dep variable, for each category.
>
T. Scott Thompson replied:
> No. If all of the data are close to being on a line, then this is
> true for any subset as well. In fact because we are now allowed to
> vary the line across groups, the within group collinearity will tend
> to be more severe. If anything the problem is worse, since you have fewer data
> points within each group, and at least as much collinearity.
Scott, thanks for your responses and clear illustrations. I understand
your points (I think). Still, I think you are envisioning a higher
correlation of the ind variables than I am. Imagine more of a data
cloud, with a general trend at a 45 degree angle away from the origin,
with an r of say 0.5. If it is clear from a scatter plot that the x
axis can be broken into regions such that the correlation between the
ind variables in each of those regions is significantly less than the
correlation over the full range of the data, I don't see why ANCOVA
wouldn't be a suitable way of estimating the independent effects of the
two ind variables on the dep variable, at several levels of the
categorized ind variable.
> A final point: You say that you realize that you shouldn't make
> forecasts that involve varying a regressor outside its observed range.
> I can't think of any argument supporting this view that doesn't also
> tell you that you shouldn't make forecasts that vary a pair of
> regressors outside the observed range for the pair. Extrapolation is
> extrapolation, whether the individual variables, considered one at a
> time, have reasonable values or not.
Ok, I see that--thanks.
