Mathematics In Finance
  
    Project and Presentation, 2000
  

   
    GUIDELINES FOR THE REVISED PRELIMINARY PROPOSAL
    
While most of the preliminary proposals were good, all of them ommitted
some aspect or what was asked for, usually in references or methodology.
Some of the proposals showed that the student had not had much contact
with the advisor, while others showed much more contact.  The advisor
is a primary resource.  You should be meeting with him/her for an hour
each week.  I would like you to revise and finalize your project proposal, 
making sure to answer the following questions:

A. What will the project accomplish.  How will the results of the project
   be used?
   
B. What kind of models will you use?  This is an important point and 
   one that requires creativity and mature judgement.  For example, 
   if you are interested in bond or loan default, you might model the
   yield spread as a stochastic process, or the credit rating as a
   stochastic process, or just think of defaults as random times
   (a Poisson process).  This has implications as to what data you
   can use, what correlations you can get, and so on.  Even if you are
   doing something statistical, like looking for correlations between
   loan prepayments and weather, the statistical tests for significance
   are model based.

C. What data will you use?  Where will you get it?  Is there enough
   data to make meaningful conclusions?
   
D. Does the reading list contain articles on what other people are 
   doing to solve your problem or closely related problems?  Generic
   math finance or statistics books do not qualify.  What do you 
   expect to learn from the references?  One of the purposes of the
   project is to learn how to read the literature.  This should be
   a significant part of your project.

A final note concerns confidentiality.  While you are working on your
project, you are free to use confidential data and produce confidential
results.  However, the eventual presentation will be public.  You need 
to find a way to produce non confidential results that illustrate what
you have done and why it is good.  This might mean running your model
on other data (e.g. domestic instead of foreign).