> There is a well known separation between the theorist and the experimentalist mindset. He was one of the best of the experimentalists.
He was a terrible experimentalist. He never used randomization or blinding; he never took into account any covariates like travel or smog (he lived in China and flew back and forth! and would still report A/B/A comparisons); he didn't even try to correct for time trends; and all of this was deliberate since he knew why you need good methodology and how abandoning all this stuff leads to systematically false results, and he went and did it all anyway. Roberts as critic was very different (and much superior to) Roberts as experimentalist.
He was a terrible experimentalist. He never used randomization or blinding; he never took into account any covariates like travel or smog (he lived in China and flew back and forth! and would still report A/B/A comparisons); he didn't even try to correct for time trends; and all of this was deliberate since he knew why you need good methodology and how abandoning all this stuff leads to systematically false results, and he went and did it all anyway. Roberts as critic was very different (and much superior to) Roberts as experimentalist.