Hacker News new | past | comments | ask | show | jobs | submit login

Correlation does not strictly mean causation. Our statistics professor brought this example once, he explained it that people who don't drink (in a western society) have often problems with their health or are taking medication. Statistics could be often misleading.

Another example was that people with bigger feet were more likely to be earn more money. Guess why?




While true, it's not just correlation that alcohol may be good for you. There's growing evidence that, among other things, alcohol boosts your HDL (good) cholesterol. For one controlled study see:

http://circ.ahajournals.org/cgi/content/full/102/19/2347

The conclusion from the controlled study?

"Alcohol intake increases HDL-C in a dose-dependent fashion, associated with and possibly caused by an increase in the TR of HDL apolipoproteins apoA-I and -II."

Of course, this is different from the author's hypotheses about socializing and loneliness, something much harder to test.


But now we are several steps removed from the outcome of interest. We can postulate the following:

EtOH -> higher HDL -> increased reverse cholesterol transport (or another HDL function) -> reduced risk of cardiovascular disease.

But there is a problem here. While higher HDL epidemiologically correlates with reduced CVD, the only clinical trial of a drug that purely raises HDL showed increased mortality. Now, the trial was complicated by off-target effects of the drug, but my point is that we actually do not know if raising "good" cholesterol is good!


You make very good points, and I think it's far from proven that alcohol actually is good for you or that raising HDL by itself is a good idea (in fact, there are some populations with low HDL but very low CVD). Probably another overlooked factor is the size of the HDL particles. Or it could be, as some people argue, that HDL is just a marker and so affecting HDL directly won't change your health.

Torcetrapib is complicated as a CETP inhibitor. The drug didn't just inhibit the conversion of HDL to LDL (which means it didn't just boost HDL, but it also lowered LDL), it also inhibits reverse cholesterol transport (where CETP is involved). Patients on torcetrapib had weird distortions in a number of blood values, increased blood pressure, and other effects. Which was the cause of increased mortality? It's too early to know, although it's pretty clear that CETP inhibitors are a bad idea.

My point was more that we have more than just correlations that people who drink alcohol have lower mortality and cardiovascular risk (observational studies often turn out to be misleading), we also have some somewhat plausible mechanisms for why this would be the case. It'll definitely be a long while before we can actually know whether or not alcohol has net positive or negative affects.


Given that this actually is my area of expertise, I'd be quite interested to hear an explanation for the downvotes.


Doesn't niacin reduce mortality?


Yes, but it is not a pure HDL modifier: it also reduces LDL-C, so it is not clear which effect leads to niacin's benefits (both could, to be fair).


Because men earn more money and tend to have bigger feet than women?


Yes. Or that's at least the professor's theory.


Could it not be because people with bigger feet tend to be taller? (I don't have evidence that big feet -> tall, but I'm pretty sure it's the case)

It's been demonstrated in the past that taller men tend to be more successful.


Because taller people tend to earn more money and people with bigger feet tend to be taller?


Er, the whole point of the article is that the health benefits might not be caused by alcohol consumption, but rather by higher degrees of socialization that happen to be correlated with alcohol consumption.


My point is that you can't explain a statistic with one cause. Most of the time there are many. Some of them we know, some of them not.


Shot in the dark, foot size is correlated with overall size, which is correlated with childhood nutrition, which is correlated with socioeconomic status




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: