Comment by vikramkr | Hacker News Reader

Comment by vikramkr | Hacker News Reader

It looks like this is the equivalent of a preprint. So, while an interesting finding, a general reminder to take it with a grain of salt and recognize that there might be issues in the statistics (maybe confounding variables that weren't controlled for) that would lead to revised results if fixed, and we've also seen peer reviewed publications like the Lancet paper that were also hit with data integrity issues. It's a good data point to keep in mind, and you should probably be wearing a mask, but we'll need a meta-analysis of a lot of these studies over the next year or so to know to any level of certainty how much they actually helped. That's not useful for our behavior right now of course, but in terms of drawing conclusions, always keep in mind that initial analyses lime this are just that, initial. From the paper:

"IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper should account for its provisional character. A revised version may be available directly from the author."

I am of the opinion that preprints should basically not be posted on HN at all. Most of the general populace of HN does not have the expertise to understand the limitations of such work (perhaps outside of CS and Math). Even physics (my area) pre-prints posts are usually filled with speculative (read wrong) theories, and misinterpretations of textbook physics.

Preprints that deal with the human world, or analyze human data are dangerously outside the expertise scope of the typical HN reader. Dangerous, because people walk away with wrong conclusion, rather than a "failure to parse" error.

You're assuming also that peer review will fix statistics issues. Statistics will often fall through the cracks.

That's why I mentioned meta-analyses as being the most important in the end. No one paper in these fields is enough to make any conclusions.

Nobody thinks the finding itself is interesting. The main point of the paper is the confirmation that simple face masks work in 3% - 10% of the cases, arguably no protection.

The new typical approach of this paper is the political angle with statistical tricks to make an insignificant number significant for the main press, by coming up with exponential growth models leading up to the cited 40% number. Which is politics, not science. But even Nature ventures into politics these days, deviating from science, with wrong growth models.

The German translation would be: "Auch Kleinvieh macht Mist" Even the smallest measures are exaggerated in desperate times. Political desperation, not scientific.

This is a completely vapid and entirely generic attack that could be posted, verbatim, in any discussion of any research paper.

I'm barely exaggerating with "verbatim": remove the parenthetical ", and you should probably be wearing a mask" as well as the quote at the end, and read it again. It could be concern-trolling any statistics paper, ever!

And what's the point, exactly? Sometimes, results from research later turn out to be wrong, and therefore it is best to completely ignore them? At some point you'll deign enough time to have passed to allow some research to be useful. But until then, you also recommend masks, because..? How is your hunch a better barometer of truth than actual research using real-world data?

It’s basically a pull request. There are going to be mistakes and changes. You can pull it and run it yourself if you’d like, but don’t say it’s shipped until it lands on master.

And it should be posted in the discussion of any research paper. Especially every preprint. Especially ones of significant public interest. Preprint sites and preprint contain the disclaimer for a reason, and everyone sharing or reporting on them should feel obligated to include a disclaimer like that because people often think science paper = fact proven when that couldn't be further from the truth. Because it's not just that things are "sometimes " proven wrong, but that you have entire fields suffering from reproducibility crises. Including "hard" science fields like oncology.

Indeed. The first response to anything challenging someone's (imprinted?) worldview is to nit pick on some detail, in order to invalidate the uncomfortable thesis with which they are presented.

This paper isn't challenging my worldview. It's supporting it. I posted the disclaimer because the paper is a preprint and it's easy to not get that fact when you just look at the title of this post and scroll through the comments.

Reply on news.ycombinator.com