Posted on November 2015 by Brian McGill
Teaching a graduate statistics class, I end up as a statistical consultant a lot. One of the questions I get most often is should I treat this as a fixed or a random effect? This topic seems to be shrouded in mystery. Indeed when I came of age statistically in the dark ages (=20 years ago), the main distinction given between a fixed and a random effect was philosophically based: are you measuring a few specific instances of interest in themselves (=fixed) or a few randomly chosen instances interesting only as representatives of a population (=random). This is not a bad approach, and seems clear to me, although I have to confess I have not had great luck teaching this distinction.
More information here