FINAL  Stat 798C  PROJECT  DUE  May 19, 2003 by 12 noon.
================================================

The Final Project, which will count as much as 2 Homework Sets, is to 
be a 5 to 10 page paper on a topic or data analysis of your choice. At least 
5 of the pages should consist of narrative, mathematical exposition and 
notation, etc., as opposed to computer output. The computer results and 
pictures you hand in should be carefully chosen and edited to support the 
points you are making in the main text.

Unlike the small data analyses and tasks you have done throughout the 
course, the purpose of this Project is for you to put together a coherent 
and reasoned train of thought, which may involve data analysis and 
graphical presentation, illustration and interpretation of a computational 
procedure, simulation, etc.

After choosing your topic, you should check with me to confirm its 
suitability. Here are some examples of suitable project topics.

		EXAMPLES OF PROJECT TOPICS

(I) You could do a Case Study or extended data analysis on a dataset 
downloaded from StatLib or other source.  Many journal articles over the 
past 15+ years have datasets which have been posted to on-line archives 
such as those on StatLib. The Project would consist of a package of 
alternative analyses, including graphical display, residuals analysis if 
appropriate, some kind of goodness of fit confirmation, etc. Your chosen 
dataset should be of sufficient size and/or complexity to support that much 
effort.

(II) There are various analyses, timing studies, simulations, etc. which you 
could present to compare Splus versus SAS or one version of Splus versus 
another (eg Splus 3.4 versus R). One example of this sort, discussed in class
and begun in my latest installment of Lecture Notes (one the web-page under
/s798c/Handouts/Lec03Pt5D.pdf) is an extended timing comparison of 
different ways to code simulation-loops in Splus involving large datasets and 
statistical analyses within each iteration. Comparisons of Splus versus R, and
of different ways to break up large loops into smaller loops and functions, 
would be suitable and interesting.

(III) You could also choose some Venables and Ripley topic not covered in 
class, write up a short summary of the statistical rationale for using it on 
certain kinds of data, code some illustrations, and show how to interpret the 
output list-components (and confirm --- perhaps via simulation --- that you 
have the interpretations right !) Three examples of such topics:
	--- gam  (Generalized Additive Models)
	--- lme, nlme  (Linear and nonlinear regression with random effects)
	--- time series, various possible methods

MANY OTHER POSSIBILITIES EXIST ! TALK TO ME ABOUT 
THEM IF YOU WANT POINTERS.