Your comment about keeping things as simple as possible is correct. This
should always be the case, but one should also take care not to
oversimplify an experiment to the point that it is meaningless.
Essentially this is what the gum-pigment ration test did. Why ... because
each of the variables depend on each other. If they didn't then the test
would in fact be valid.
Designing experiments to deal with correlated or depedendent variables is
simply not an easy thing to do.
- Wayde
(wallen@boulder.nist.gov)