There are many cases where a linear projection is made, probably because the
plot implies as much to the viewer's eye, not because there is a physical reason
to believe in it.

the total human population of the Earth is one.  Try projecting the population
between 1990 to 2000 on into 2100 or 2300 CE.  It gets worse if you pick one
country during a period of industrialization & curtailment of communicable
disease, and project on into the future.

Or take the sales of computers, desktop variety, between 1985 and 1995.  Project
that forward linearly, as so many people did.

there is usually no reason, a priori, why a time series should continue a linear
pattern indefintely.  Where does the passage of time _cause_ a change in the
measured response?  (BTW, I refuse to get started on a discussion of causality
:)

Because someone uses a linear model is not reason to damn the procedure.
Neither should we believe the projection as a forecast, simply because someone
thought a straight line fit of data was a good idea.  It was more likely the
best one they could handle, and they wanted to believe the result.

I vote that we accept "projection" as a mathematical extrapolation based on a
model of system behavior, and usually a "simple" linear projection.  A
projection may or may not be in a time direction.

And that we accept "forecast" as an estimate of behavior, always extended into a
(time direction) future period.  Said estimate based upon a complex model of
system behavior which may or may not include only strict mathematical
calculation.  That is, it includes synthesis or higher levels of understanding
as described in the Bloom taxonomy.

When I write these descriptions out as given here, I see that the difference is
mostly one of degree, not a categorical distinction.

Time to get back to work.

Cheers,
Jay

Sandy MacRae wrote:

> In the argument by Paul Garcia:
>
> > Take the set of all projections: they are all forecasts, but with no extra
> > dimensions added.  But it is possible to produce forecasts which are not
> > projections, but based on other stuff (prophecies, gut feeling, etc.).  So
> > the set of forecasts is bigger than the set of projections and includes all
> > the projections.  Ergo: projections are a subset of forecasts.
>
> The first statement is not incontrovertible and I, for one, do not agree
> with it. Some projections are untenable as forecasts because they are
> contradicted by other evidence.  If in successive months a baby has 0 ,
> 1, 2, 3 and 4 teeth, it would be a projection to predict 60 teeth after 5
> years but I would not offer it as a forecast.
>
> Thus, as Dave Stewardson already said, neither 'projection' nor
> 'forecast'  is a subset of the other.
>
> Sandy MacRae
>
> StatBasics
> 6, Ascot Road
> Moseley
> Birmingham
> B13 9EL
>
> Phone / Fax: 0121 247 6138
>      http://www.statbasics.co.uk
>      http://www.difftest.co.uk

--
Jay Warner
Principal Scientist
Warner Consulting, Inc.
4444 North Green Bay Road
Racine, WI 53404-1216
USA

Ph: (262) 634-9100
FAX: (262) 681-1133
email: [log in to unmask]
web: http://www.a2q.com

The A2Q Method (tm) -- What do you want to improve today?