I.  Procedures.

1.  From the literature (journals, books, monographs,
proceedings volumes) find an example of econometric work on
an interesting topic.  I am not suggesting you should
directly replicate a previous study - you will often wish to
simplify or otherwise modify the approach.  You can use ad
hoc theory to specify an equation, but it does not usually
work very well.  In any event, talk to me about what you are
doing, especially if you do not have a "guide" from the
literature.

2.  Obtain you data from wherever you can get it - the
internet, Statcan, etc.  I recommend machine-readable
sources unless you are good at data entry, though you may be
able to scan in printed data and use OCR (optical character
recognition).  See me if you have interesting data that is
not machine readable.  I suspect most people will use
CANSIM, but SFU has a number of other data sets available.
In general, use as many observations as you can get without
prohibitive cost (unless you are using large micro data
sets).

3.  After you get your data, plot the series to check for
missing observations and other errors, and to "see what you
have".  After estimating the equation or equations
comprising your project, check for nonlinearities,
structural breaks, heteroscedasticity, autocorrelation,
multicollinearity and any other econometric problems you
suspect may be important, and correct as necessary.
Autocorrelation should not be considered if your
data are cross-sectional.  Even though time series data
seldom display heteroscedasticity, you should check for
this.  If your data are quarterly or monthly, you should
check for seasonality and include dummies if needed.

2.  In general I suggest you do tests which will affect
specification (testing for seasonality, structural breaks,
nonlinearity) and corrections based on these tests before
correcting for heteroscedasticity and autocorrelation with
EGLS.  Note that the presence of heteroscedasticity and
autocorrelation is often evidence of specification error.
Do RESET and other specification error tests.

3.  You may decide to go back and get more data (either more
observations or more variables) on the basis of what you
find out in step 3 above.  This is not required if you did a
reasonable job in steps 1 and 2, and you will in general not
have time to do much of this anyway.

4.  Although these projects will be graded, I am very
willing to help with questions about how to proceed, how to
interpret results, technical details about SHAZAM usage, and
so forth.  I am much more concerned about what you learn in
the process of working on the project than I am about
grading.

II.  Turn in for grading:

(1)  A statement of what you were trying to do (e.g.
replicate some study, or estimate some parameter(s) or test
some hypothesis).

(2)  A xerox of the relevant portion of the previous
research (e.g. article) that most affected your
specification.  Two or three pages showing previous
empirical results should suffice.

(3)  A brief statement of where you got your data (CANSIM
numbers if that was your source) and description of
transformations and adjustments made (splicing series,
coverting monthly to quarterly, etc.).  Note any problems
you had obtaining data (e.g. you couldn't find the series
you really wanted, but you found a useful proxy variable).

(4) A short (3-4 pages, if you can be concise) chronology of
what you did and what you concluded.  For example, "I first
added quadratic terms to all variables to test for
nonlinearity, and finding none, I added seasonal dummies.
Going back and testing for nonlinearity with the seasonal
dummies included I now found nonlinearity involving X2.  I
then ...".

(5) A copy of your SHAZAM command file (if you work
interactively, a reasonably complete hard copy output).

(6)  As much output as needed to illustrate points raised in
your writeup in (4) above.  Cross reference equations or
results in your output to statements in your writeup or
command file (highlighted or pencilled notations work well).
It is generally a waste of time to retype equations into
your writeup.  This is not a term paper.