-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On
Behalf Of Herman Rubin
Sent: Friday, April 05, 2002 12:53 PM
To: [EMAIL PROTECTED]
Subject: Text for numerical methods in statistics
I am looking for a good text for such a course; I have
taught this many times, but I have been dissatisfied
with the texts I have used, not that I am likely to
really like another one.
The aim of this course is not to carry out routine
analyses, but to convey the ideas needed to find good
methods for computing in non-routine situations. The
level of the course is that the student has had a
graduate level course in probability, and also in
statistical theory, and is on speaking terms with
linear algebra, real analysis, and complex analysis,
but need not have had any previous course in numerical
analysis; I do not believe in cookbook before theory.
-------------------------------------------------------------------------
I've read your message, the message from B.D. McCullough, went to Monahan's
web page, went over the table of contents and went through several Fortran
programs.
What is your objective? Do you want to have students just know what black
boxes to haul out to work on a problem or do you want your students to
understand the logic within the black boxes. FORTRAN is not a good language
to understand what is going on. All the heavily nested subroutine calls and
the obscure input/output make it almost impossible to understand what's
going on.
Do you want to train programer's on how to develop commercial, copyrighted
stat programs?
If you intend to try and impart understanding of "how things work", G.W.
Stewart's books and methods is a good way to go. He uses an algorithm
shorthand, from which he can discuss what is going on, step by step. From
the algorithms, one can write actual programs in C++. VB, Java, Fortran,
etc.
If you want to teach black boxes using SAS or S or SPSS or LISREL.... that
is one way.
If what you want to teach is how to take a JASA article (i.e. buried in
math) and how develop a computer program/subroutine/module ..... that does
the computations, then you would have to show how the math is interpreted in
language constructs.
I am a little apprehensive about a book from Cambridge Press dealing with
computer programs. They were the ones who put out "Numerical Recipes", a
notorious collection of bad and buggy programs, and refused to put out a
corrected edition.
DAHeiser
PS. Fortran was my first computer language back in 1960 on the old GE 225
and IBM 7480? with magnetic bead memory.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
