4 ms·

I really like this paper. Why? Because it is a randomized controlled trail. There are not enough of these in the medical field. This is the gold standard for sc

I really like this paper. Why? Because it is a randomized controlled trail. There are not enough of these in the medical field. This is the gold standard for scientific research. Unlike in the comment section of most health related science articles, there will now be no discussion of "correlation is not causation!!!" because the subjects were chosen at random. Additionally the subjects were studied genetically and the results were used to divide the results into two groups: those who metabolize alcohol quickly and slowly. This allows for much more understandable results because it explains why the treatment does not work in some cases or works well in others. The results of the study are not earth shattering, however, it is able to come to a much more firm conclusion that the typical epidemiological study that relies on statistically examining the population at large. Those types of studies are great for figuring out what to study next. Notice some correlation (even corrected for all the biases that you could come up with and measure) that people who eat more saturated fat have higher chances of heart disease? Then do not change the diet recommendation for all Americans. Instead, do a randomized controlled trail! Then make your recommendations.

Darmani 11y ago
I can't tell if this is sarcasm. RCTs are incredibly commonplace in medicine, but they are hardly the only way to do causal inference. They are also incredibly easy to make mistakes in; I'd more call them a bronze standard.
- gwern 11y ago
  They're not a commonplace in the eternal debate over whether moderate alcohol is healthy or not, which is why this paper is so awesome. The other ways to do causal inference are also much more unreliable; I'll take a questionable RCT over a Mendelian randomization or regression-discontinuity design any day.

Darmani 11y ago

I can't tell if this is sarcasm. RCTs are incredibly commonplace in medicine, but they are hardly the only way to do causal inference. They are also incredibly easy to make mistakes in; I'd more call them a bronze standard.

gwern 11y ago
They're not a commonplace in the eternal debate over whether moderate alcohol is healthy or not, which is why this paper is so awesome. The other ways to do causal inference are also much more unreliable; I'll take a questionable RCT over a Mendelian randomization or regression-discontinuity design any day.