ecological fallacy

[mcnaught@Lanl.GOV (Mike McNaughton)]

I would be grateful for your comments on the following idea.

Bernie Cohen's studies of the correlation of lung cancer with dose from
radon have been criticized because they are limited by the "ecological
fallacy", i.e. that average doses do not necessarily determine average
effects.  This same mathematical problem occurs in many places in physics
(e.g. in the design of nuclear weapons) and is addressed by "Monte Carlo"
calculations.  I propose to write a Monte Carlo computer program for this
purpose.  The program would replace the "averages" used in an ecological
study with "simulated individuals".  Variables would correspond to radon
dose, smoking prevalence, socio-economic variables, etc, and would be
distributed according to census data.

The benefits might include the following.

1. Ecological studies are the only type of study with the required
statistical precision.  A Monte Carlo calculation would help to establish
ecological studies as mathematically valid methods.

2. This calculation would place further constraints on the possible shape
of the dose response curve.

3. If this computer simulation program is disseminated via the internet, it
could be used by anyone, and help to make people more familiar and
comfortable with the ideas, in much the same way as computer simulations
help in undergraduate physics.

If you are interested I would be glad to discuss this proposal in as much
detail as you wish.

Thank you, mike (mcnaught@LANL.gov)

Stay safe, mike (mcnaught@LANL.GOV)