a better example:

drinking coffee *causes* alertness during exam period.
bottled water does not affect alertness.

.'. the true life equation is:
alertness=coffee

but say we forget to measure coffee in our model but feel
exploratative,
the fictional model is:
alertness=bottled water+body weight+age


I assume that the only way to know the validity of the coeffiients
(and their direction!) is to look at how much of the ultimate variable
is predicted.
[this seems unreliable given small overall predictive power of
regression models seen in exploratory research - ex. does egg
consumption correlate with serum cholesterol?]
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to