Hacker News Reader

Comment by soperj | Hacker News Reader

Somehow all the studies end up being on jewish kids in Isreal vs. jewish kids in England. Not exactly a representative sample of people across the globe. The studies themselves have all the hallmarks of shit science, but everyone on here seems to agree with it because it fits their world view.

View on HN

Well the opposite view isn’t really based on science , only in faulty risk mitigation. So is it riskier to let your kid go without knowing for a long time or is it better to test in a limited controlled window where you can react and adapt if something goes wrong.

Neither is based on Science when the science is shit. P-hacking isn't science, and yet it gets published as Science all the time, particularly in nutrition studies.

But that's exactly why this is so interesting. You take a set of people with extremely similar genetics but somehow have very different reactions.

I suppose a technical (albeit not precise) analogy would be you producing 100 servers and selling half to both AWS and Azure. Then you find out that 25 of the servers at Azure are overheating and dying while AWS doesn't seem to have this problem at all. So you look into it and find that AWS is air-cooling their systems while Azure isn't. Sure you could argue post post hoc ergo propter hoc but you'd also probably call the folks at Azure and tell them to install an air-cooling system.

Extrapolating that to your point, you're saying it would be "shit" to tell the folks at Azure to install the air-cooling because you didn't test this acrossa majority of servers you've produced. The folks at Azure would point the incredible similarities (this is the same server model, the same chips, the same manufacturing run etc.) and then would promptly find another vendor.

> But that's exactly why this is so interesting. You take a set of people with extremely similar genetics but somehow have very different reactions.

Or it's based on a hundred different other things than when they first tried peanut butter.

Your point of anything is possible is always and forever true.

The point here is that there seems to be a strong correlation between people eating Bamba and no peanut allergy. How can we develop a control group? Oh wait, there is a group of people that have the same genetics but doesn't consistently eat this...perhaps there's a connection.

The goal is to see if we can show a correlation/causation.

As always, correlation is not causation, and when you have a large group of people who actually changes their behaviour and starts to feed their children peanut butter early and it has no effect whatsoever you can begin to rule out the theory that was based on a small sample size of a distinct population.

The entire point was looking at very similar populations, but controlling for one variable (peanut product consumption in early childhood). While I agree that the studies should be expanded outward to ensure reproducibility and verify the hypothesis, a useful starting point is to look at two similar populations with only one variable.

That original study was made 8 or 9 years ago now, and they're following it up with an additional study of jewish kids in britain and israel here. If you look at the study, they're trying to determine why early exposure to peanut butter has not decreased the incidence of allergy in Australia. Somehow a study of Jewish kids in Israel & Britain is supposed to prove this out.

Reply on news.ycombinator.com