4 ms·

> machine learning? Not sure why you assumed this. ML isn't the only "innovative statistical method". I haven't read the paper though. > "depression in early

> machine learning? Not sure why you assumed this. ML isn't the only "innovative statistical method". I haven't read the paper though. > "depression in early adulthood is correlated with dementia later in life" That is how all inferences are made. It is very _very_ rare to be able to demonstrate strict causation in medical sciences. This is why the title says "may".

joconde 5y ago
> machine learning? >> Not sure why you assumed this. ML isn't the only "innovative statistical method". I haven't read the paper though. I have to assume things since I don't seem to have access to the actual paper. "Innovative" methods that predict things made me think of machine learning, but I'd be happy to learn more if I could read the paper. > It is very _very_ rare to be able to demonstrate strict causation in medical sciences. This is why the title says "may". The title is a lie. It very clearly suggests that the study gives some evidence of a causation, while the method and results don't support that interpretation at all. If whoever wrote this wanted to honestly inform people about the result, they could have written "people who are depressed early in adulthood have higher chance of dementia later in life", which makes it clear for someone paying attention that this doesn't exclude a confounding factor.
- thucydides 5y ago
  I agree with your first post but to call the title a lie is not right. It seems to me that “happiness in early adulthood may protect against dementia” is logically consistent with your proposed rewrite, “people who are depressed early in adulthood have higher chance of dementia” On reading the title alone, I assumed a university marketer was trying to reframe in an interesting way what had been found in the study (something akin to your proposed summary)
  - joconde 5y ago
    > It seems to me that “happiness in early adulthood may protect against dementia” is logically consistent with your proposed rewrite, “people who are depressed early in adulthood have higher chance of dementia” Saying "the moon is white" is logically consistent with saying "the moon may be made of white cheese". Logical consistence isn't an excuse for a major university to make people believe that research supports something that isn't actually supported by their work. > On reading the title alone, I assumed a university marketer was trying to reframe in an interesting way what had been found in the study (something akin to your proposed summary) Yes that's definitely the reason. I couldn't find the article on the journal's website, but I imagine that the paper was honest about this being a correlation. We can debate whether the title (and some of the article) fits the definition of a lie, but it's 100% clickbait that will make some people believe something that has no basis in research, and they are exploiting that for PR reasons. And I actually wouldn't be surprised if the causality they suggest is true... But that kind of exaggeration makes it impossible to follow what we actually have evidence for, which is the opposite of what this kind of popular science articles should be doing.
  - therealcamino 5y ago
    To me it's not the same statement: it's one thing to say that people who might be prone to be depressed as early adults might have higher risk of dementia later. To say that happiness at that period is protective, suggests that if you cheered them up at the right time, they'd no longer be at higher risk. Is it something about their brain structure or approach to adversity, or is it specific, moment to moment lived experience? If it's the latter, then a dementia prevention strategy could be giving out free amusement park tickets to 20-somethings. I'm being facetious, but the "happiness" title really does shift the focus. So I didn't think it was equivalent. As a general statement, these university press releases on research findings are some of the lowest-quality posts on HN.
potatoman22 5y ago
What other statistical techniques can predict future outcomes while controlling for confounders?
JamesBarney 5y ago
But the path from "we found a correlation" to "we have good evidence of causation" is a very long path where 99% of hypotheses drop off. And this one is especially suspect. It's possible that early life happiness is protective against dementia but mid and late life less so, but our model of biology says that's less likely because almost all similar findings that we have found to be casual recency increases the effect. For instances what you eat and how you exercise this year impacts your current health more than what you ate and how you exercised 2/5/10/15/20 years ago. I think "may protect" is misleading because while it might be the case that if you increase someone's happiness in early adulthood it will decrease their risk of dementia, it's certainly not likely. For instance if you found some depression RCT from 50 years ago in 20 somethings, what are the chances the control group and the intervention group are going to have significantly different dementia outcomes? So while it's technically true that it "may protect against dementia", it was also technically true before someone did this study. And honestly studies like this have so many degrees of freedom they are almost always unintentionally p-hacked.

joconde 5y ago

> machine learning? >> Not sure why you assumed this. ML isn't the only "innovative statistical method". I haven't read the paper though. I have to assume things since I don't seem to have access to the actual paper. "Innovative" methods that predict things made me think of machine learning, but I'd be happy to learn more if I could read the paper. > It is very _very_ rare to be able to demonstrate strict causation in medical sciences. This is why the title says "may". The title is a lie. It very clearly suggests that the study gives some evidence of a causation, while the method and results don't support that interpretation at all. If whoever wrote this wanted to honestly inform people about the result, they could have written "people who are depressed early in adulthood have higher chance of dementia later in life", which makes it clear for someone paying attention that this doesn't exclude a confounding factor.

thucydides 5y ago
I agree with your first post but to call the title a lie is not right. It seems to me that “happiness in early adulthood may protect against dementia” is logically consistent with your proposed rewrite, “people who are depressed early in adulthood have higher chance of dementia” On reading the title alone, I assumed a university marketer was trying to reframe in an interesting way what had been found in the study (something akin to your proposed summary)
- joconde 5y ago
  > It seems to me that “happiness in early adulthood may protect against dementia” is logically consistent with your proposed rewrite, “people who are depressed early in adulthood have higher chance of dementia” Saying "the moon is white" is logically consistent with saying "the moon may be made of white cheese". Logical consistence isn't an excuse for a major university to make people believe that research supports something that isn't actually supported by their work. > On reading the title alone, I assumed a university marketer was trying to reframe in an interesting way what had been found in the study (something akin to your proposed summary) Yes that's definitely the reason. I couldn't find the article on the journal's website, but I imagine that the paper was honest about this being a correlation. We can debate whether the title (and some of the article) fits the definition of a lie, but it's 100% clickbait that will make some people believe something that has no basis in research, and they are exploiting that for PR reasons. And I actually wouldn't be surprised if the causality they suggest is true... But that kind of exaggeration makes it impossible to follow what we actually have evidence for, which is the opposite of what this kind of popular science articles should be doing.
- therealcamino 5y ago
  To me it's not the same statement: it's one thing to say that people who might be prone to be depressed as early adults might have higher risk of dementia later. To say that happiness at that period is protective, suggests that if you cheered them up at the right time, they'd no longer be at higher risk. Is it something about their brain structure or approach to adversity, or is it specific, moment to moment lived experience? If it's the latter, then a dementia prevention strategy could be giving out free amusement park tickets to 20-somethings. I'm being facetious, but the "happiness" title really does shift the focus. So I didn't think it was equivalent. As a general statement, these university press releases on research findings are some of the lowest-quality posts on HN.

potatoman22 5y ago

What other statistical techniques can predict future outcomes while controlling for confounders?

JamesBarney 5y ago

But the path from "we found a correlation" to "we have good evidence of causation" is a very long path where 99% of hypotheses drop off. And this one is especially suspect. It's possible that early life happiness is protective against dementia but mid and late life less so, but our model of biology says that's less likely because almost all similar findings that we have found to be casual recency increases the effect. For instances what you eat and how you exercise this year impacts your current health more than what you ate and how you exercised 2/5/10/15/20 years ago. I think "may protect" is misleading because while it might be the case that if you increase someone's happiness in early adulthood it will decrease their risk of dementia, it's certainly not likely. For instance if you found some depression RCT from 50 years ago in 20 somethings, what are the chances the control group and the intervention group are going to have significantly different dementia outcomes? So while it's technically true that it "may protect against dementia", it was also technically true before someone did this study. And honestly studies like this have so many degrees of freedom they are almost always unintentionally p-hacked.