There is science which involves studying objective, physical things that have been repeatedly evaluated by independent scientists and proven time and time again.
And there is this new class of 'science' which is tantamount to p-value hacking and only slightly better than social engineering.
Coffee cures cancer. Year 2010
Coffee doesn't cure cancer. Year 2011
Black chocolate prevents heart attacks... And so on
Most of these studies read something like this: 'we asked a hundred people what they ate for the last month'. I don't even remember what I ate for my evening snack yesterday.
It's really poorly controlled and definitely not the rigor that goes with proper science. This should not even be taught in schools let alone treated as 'research'.
To be fair though, in this case the results are still the same when correcting for the methodological errors that happened in the study.
From TFA:
> It turns out approximately 14 percent of the more than 7,400 study participants hadn't been assigned randomly to either the Mediterranean diet or a low-fat one. When couples joined the study together, both had been picked to follow the same diet. At one of the 11 participating study sites, the lead investigator had assigned the same diet to an entire village and didn't tell the rest of the investigators.
> "This affected only a small part of the trial," says Martínez González. When the researchers reanalyzed the data excluding the nonrandomized people, the results were the same, he adds.
> Still, because everybody wasn't randomly assigned to different groups, the study can no longer claim the diet directly caused those health benefits. "We need to tone down the results, but it is just a little bit," he says.
> Many other studies have shown that people eating the Mediterranean diet have lower risks of various ailments, without claiming cause and effect.
In this case, since it wasn't fully random, they weakened their claim to remove the cause-effect link, instead of p hacking as you mentioned.
"The revised paper says only that people eating the Mediterranean diet had fewer strokes and heart attacks, not, as the original paper claimed, that the diet was the direct cause of those health benefits."
A typical wording for that sort of research, not really a retraction.
5 comments
[ 1.9 ms ] story [ 22.1 ms ] threadAnd there is this new class of 'science' which is tantamount to p-value hacking and only slightly better than social engineering.
Coffee cures cancer. Year 2010 Coffee doesn't cure cancer. Year 2011 Black chocolate prevents heart attacks... And so on
Most of these studies read something like this: 'we asked a hundred people what they ate for the last month'. I don't even remember what I ate for my evening snack yesterday.
It's really poorly controlled and definitely not the rigor that goes with proper science. This should not even be taught in schools let alone treated as 'research'.
From TFA:
> It turns out approximately 14 percent of the more than 7,400 study participants hadn't been assigned randomly to either the Mediterranean diet or a low-fat one. When couples joined the study together, both had been picked to follow the same diet. At one of the 11 participating study sites, the lead investigator had assigned the same diet to an entire village and didn't tell the rest of the investigators.
> "This affected only a small part of the trial," says Martínez González. When the researchers reanalyzed the data excluding the nonrandomized people, the results were the same, he adds.
> Still, because everybody wasn't randomly assigned to different groups, the study can no longer claim the diet directly caused those health benefits. "We need to tone down the results, but it is just a little bit," he says.
> Many other studies have shown that people eating the Mediterranean diet have lower risks of various ailments, without claiming cause and effect.
In this case, since it wasn't fully random, they weakened their claim to remove the cause-effect link, instead of p hacking as you mentioned.
A typical wording for that sort of research, not really a retraction.