Statistical Regression of Time Series.(Was: Sabbatical...)

[brian whatcott <inet@intellisys.net>]

At 13:16 5/20/98 -0400, you wrote:
> ... The American Association of
> University Professors has had an immense effect
> on salaries, total compensation and working
> conditions. A historial review will dramatically
> demonstrate a strong correlation between the
> development of the AAUP and the improvement in
> working conditions.
> ....
> Don Kniffen
> Hampden-Sydney College
I take this quote of Don's as a very convenient springboard
for mentioning the excellent power of statistical regression 
of time series in demonstrating relations between formerly 
unsuspected factors: see for example the excellent piece of
epidemiology involved in showing the link between small particle 
(<10 um) pollution and mortality, as discussed in the current 
Scientific American.

 This work attracted the MacArthur "genius" award for a government (!!)
scientist. Confounding factors are flattened by the time series 
correlations: they do not vary like the test factor, and that can 
obviate very expensive blind 'balanced' trials.

  The principal tools mentioned in this outstanding account were
1) a statistical regression package with non-linear features.
2) Access to a long British time series implicating sulfur 
dioxide.

The subject was a person who could not find a job as a physicist
 earlier in his career. His results have triggered US legislation 
which is thought to have saved many lives already.

The materials I mentioned above would cost any of you approximately 
zero to obtain.

 (Sorry for the scanty reportage - my copy is at the office...
I am STILL waiting for a Nobel to be awarded to one of this list's 
readership: but a MacArthur would do nicely!)

Brian Whatcott    Altus OK