Hacker News new | ask | show | jobs
by lathyrus_long 1483 days ago
The p values are small enough that the correlation seems likely to be real, even after lots of dredging through that dataset. I'm less sure about the causality, though. It's also possible they've just discovered that richer people (a) live longer, and (b) like organic produce. They controlled for various factors (note 1 on table 2), but none that seem like usefully strong proxies for income or wealth.
1 comments
minsc_and_boo 1482 days ago
The study doesn't seem to claim causality, just that this is their second study showing the same results, and that further study is needed.

I'm not sure how controlling for income/wealth would matter, given that the study is all nurses and they control by ethnicity.

link
lathyrus_long 1480 days ago
Surely this relationship is noteworthy only to the extent it's potentially a causality? The authors are careful not to make explicit claims, but most discussion here is about that possibility. The title of this HN submission is too ("could negate"), though the title of the paper itself is more cautious.

After more carefully reading the paper, I think the premise of my original comment was completely wrong though. The study apparently had no information on whether the participants were eating conventional or organic produce. Rather, they categorized by type of fruit or vegetable, converting the average results of tests for residues to a score from 0 (best) to 6 (worst) with a somewhat arbitrary heuristic. For example, grapefruit scored 0, and spinach scored 6.

That eliminates the spurious correlation I speculated about, though others may exist; it would be interesting to know which specific fruits and vegetables accounted for most of the effect. That heuristic also increases the risk of data dredging; I wonder how many versions they tried before they got their result. As to the residues, it seems they considered all pesticides equally, even though some are strongly suspected to be much more dangerous than others. It would be interesting to repeat the analysis considering only residues of those suspected to be most (or least) dangerous, to see if their effect gets bigger (or smaller).

link