ABSTRACT
As an aside, I remark that in one of my earliest publications (McDonald & O’Hara, 1964), on size-distance invariance, I had an eight-factor split-split plot
design, which gave me a technically significant seventh-order interaction involving apparatus-counterbalancing conditions, and some huge desired main effects. I cheated by reporting individual tests on the main effects. I have never lost sleep over this, but 30 years later I possibly am inventing excuses. A central motive and a central focus for the following remarks is the problem of ignoring a statistically significant interaction in order to interpret main effects.
