3 ms·

I'll save everyone a click: yes, this is a case of 'correlation != causation'. The authors (whose freshman stats classes were presumably as good as yours) were

I'll save everyone a click: yes, this is a case of 'correlation != causation'. The authors (whose freshman stats classes were presumably as good as yours) were evidently well aware of this distinction - which is why they took the trouble to say 'association' rather than 'causation' in the title. So the pedantry here seems rather misplaced.

gwern 3y ago
No, my pedantry is not misplaced. Let me again repeat: we know the correlate is confounded. This isn't a hypothetical. It is a known fact established from many different methods triangulating the same result: the naive SCZ/marijuana correlation is confounded, and quite substantially so. Their entire analysis is predicated on an assumption which is substantially false. And I agree they know what they're doing, which is why it makes me so angry to read their paper and I am criticizing them personally and not simply raising a technical point: they know they are peddling bullshit, and at best, they want to believe, and at worst, they are cynically manufacturing scary media-friendly results they know will be cited more than the results they should have published, 'the SCZ/marijuana relationship was attenuated by over half when we ran the standard within-family analysis on the population register like we should have, yet again demonstrating how confounded it is and how hard it will be to infer any causality, sad to say for those looking for a quick hit of scary controversy and reason to ban marijuana'. The authors themselves cover up and ignore the many past results demonstrating confounding, fail to do better when they so easily could have, and then have the sheer unmitigated chutzpah, the incredible gal, to just say "assuming causality" right in the abstract to immediately start advertising huge inflammatory claims like "one-fifth of cases of schizophrenia among young males might be prevented by averting CUD." (again, from the abstract, and repeated throughout). Yes, and 'assuming the moon is made of cheese', we could save a ton on moon base supplies and might avert one-fifth of moon base costs by mining the cheese... But I don't advise going to the media about your exciting new research about how NASA could trim its budgets. I don't blame the people reporting on this for getting it wrong, when the authors spin it so hard that a CD would explode.
- lisasays 3y ago
  Lemme give the paper a closer read + get back to you on this.
- lisasays 3y ago
  Okay then - I agree with you that wording in the article seems rather shifty on this point. Like the way they start off in the title using neutral language, but then throughout the article they do seem to be spinning in such a way that it seems they want the reader to feel that in fact the relation is (or "may be") mostly causative. And that, for not addressing the elephant in the room (the correlative confound) overall the article can be said to be somewhat disappointing. What I'm not so sure about is to what extent we can, starting from that -- make inferences about the authors' broader motives, or to speak of their "gall" and how unmitigated it was. And frankly, I don't particularly care. Not that you shouldn't care, though. But if you want to criticize them "personally", perhaps a letter to the original journal would do the trick? Or at least get it it of your system? Because I kind of doubt they're taking the trouble to dig into the fine print of this thread, or to otherwise allow their morning coffee ritual to be upended by this withering analysis of yours.
- renonce 3y ago
  I wonder if this is the norm of scientific research nowadays. I mean, not the top 20% of useful researches but the bottom 80% that do not turn out to be useful immediately. We have too many scientists hungry for publications and citations so they can get funding and make a living, which means they have an incentive that distracts them from actual research.

gwern 3y ago

No, my pedantry is not misplaced. Let me again repeat: we know the correlate is confounded. This isn't a hypothetical. It is a known fact established from many different methods triangulating the same result: the naive SCZ/marijuana correlation is confounded, and quite substantially so. Their entire analysis is predicated on an assumption which is substantially false. And I agree they know what they're doing, which is why it makes me so angry to read their paper and I am criticizing them personally and not simply raising a technical point: they know they are peddling bullshit, and at best, they want to believe, and at worst, they are cynically manufacturing scary media-friendly results they know will be cited more than the results they should have published, 'the SCZ/marijuana relationship was attenuated by over half when we ran the standard within-family analysis on the population register like we should have, yet again demonstrating how confounded it is and how hard it will be to infer any causality, sad to say for those looking for a quick hit of scary controversy and reason to ban marijuana'. The authors themselves cover up and ignore the many past results demonstrating confounding, fail to do better when they so easily could have, and then have the sheer unmitigated chutzpah, the incredible gal, to just say "assuming causality" right in the abstract to immediately start advertising huge inflammatory claims like "one-fifth of cases of schizophrenia among young males might be prevented by averting CUD." (again, from the abstract, and repeated throughout). Yes, and 'assuming the moon is made of cheese', we could save a ton on moon base supplies and might avert one-fifth of moon base costs by mining the cheese... But I don't advise going to the media about your exciting new research about how NASA could trim its budgets. I don't blame the people reporting on this for getting it wrong, when the authors spin it so hard that a CD would explode.

lisasays 3y ago
Lemme give the paper a closer read + get back to you on this.
lisasays 3y ago
Okay then - I agree with you that wording in the article seems rather shifty on this point. Like the way they start off in the title using neutral language, but then throughout the article they do seem to be spinning in such a way that it seems they want the reader to feel that in fact the relation is (or "may be") mostly causative. And that, for not addressing the elephant in the room (the correlative confound) overall the article can be said to be somewhat disappointing. What I'm not so sure about is to what extent we can, starting from that -- make inferences about the authors' broader motives, or to speak of their "gall" and how unmitigated it was. And frankly, I don't particularly care. Not that you shouldn't care, though. But if you want to criticize them "personally", perhaps a letter to the original journal would do the trick? Or at least get it it of your system? Because I kind of doubt they're taking the trouble to dig into the fine print of this thread, or to otherwise allow their morning coffee ritual to be upended by this withering analysis of yours.
renonce 3y ago
I wonder if this is the norm of scientific research nowadays. I mean, not the top 20% of useful researches but the bottom 80% that do not turn out to be useful immediately. We have too many scientists hungry for publications and citations so they can get funding and make a living, which means they have an incentive that distracts them from actual research.