[Aurornis]

I'm all for installing air filters in classrooms for a number of reasons, but I also think the extreme results from this study aren't going to hold up to further research.

From the paper:

> To do so, I leverage a unique setting arising from the largest gas leak in United States history, whereby the offending gas company installed air filters in every classroom, office and common area for all schools within five miles of the leak (but not beyond). This variation allows me to compare student achievement in schools receiving air filters relative to those that did not using a spatial regression discontinuity design.

In other words, the paper looked at test scores at different schools in different areas on different years and assumed that the only change was the air filters. Anyone who has worked with school kids knows that the variations between classes from year to year can be extreme, as can differences produced by different teachers or even school policies.

Again, I think air filtration is great indoors, but expecting test scores to improve dramatically like this is not realistic. This feels like another extremely exaggerated health claim, like past claims made about fish oil supplements. Fish oil was briefly thought to have extreme positive health benefits from a number of very small studies like this, but as sample sizes became larger and studies became higher quality, most of the beneficial effects disappeared.

[AngryData]

I would also think there is a bias of schools spending extra money and effort towards student improvements if they are willing to go so far as allocate funding for air filtration systems. A better method would be to just give some schools air filtration for free and see if things help without any other major changes.

To me its like looking at schools that buy newer buses and trying to show new buses improve test scores. When in practice the only schools that are buying any significant number of new buses have far more money coming in than in the past and have a lot more to spend on students compared to other schools, which is way more relevant than what year a kid's bus is made. Maybe better buses would improve scores too, but there is no way to tell if 95% of an improvement is due to other unrelated factors based on funding.

[Aurornis]

> A better method would be to just give some schools air filtration for free and see if things help without any other major changes.

This is what the study looked at.

The problem was that it wasn’t randomized within schools or across teachers. They also looked at a very limited time window. They also note that some teachers weren’t using the air filters. They also found that the VOCs they were trying to filter weren’t even detected before the filters were used. They also used some questionable regressions to imply larger trends.

The list of problems goes on and on. It’s fascinating how easily people are tricked into pivoting around this one study, though, simply because it’s the one introduced by the headline.

[jancsika]

> They also looked at a very limited time window. They also note that some teachers weren’t using the air filters.

To correct and clarify:

They also looked at a very limited time window. They also note that-- after this very limited time window-- some teachers stopped using the air filters. In the words of the authors, this made long-run results "difficult to interpret."

[stdbrouw]

I would also expect the estimated magnitude of the effect to go down over time, but that's just my general attitude to these kinds of things, the fact is that the discontinuity design that they use already accounts for variations between classes, teachers, schools, years. The way it works is that some unexpected event that applies to some people but not others is taken to represent a natural experiment, and then variation between groups before the event is compared to variation between groups after the event. The comparison is never against no variation.

The smoking gun is really in Table 3 and Table 4, where you can see that the effects that were observed are compatible with a population effect of 0, or alternatively you can look at Figure 2 and note that you could draw a straight line (no effect) within the confidence bands. Doesn't mean the effect is not there, but that there's insufficient evidence that it is, and that we should indeed be very careful about taking the estimates at face value.

[mmooss]

Based on your comment, the effect could be larger as well as smaller.

All research is met on HN by people who know better and will tell you why it's flawed. There isn't a greater collection of expertise in the history of the world than on HN.

Edit: I meant to add: What value can we find in this research? It wasn't published as scripture, the perfect answer to all our problems. It's one study of some interesting events and data; what can we get out of it?

[Aurornis]

> Based on your comment, the effect could be larger as well as smaller.

The reality of any underpowered study could always be “larger as well as smaller”. This statement doesn’t add anything to the conversation.

The mistake is pivoting around poorly structured and underpowered research.

> All research is met on HN by people who know better and will tell you why it's flawed.

This is a misunderstanding. People who know how to read studies will always be aware of the limitations.

There’s a difference between saying “everything is flawed” and pointing out the limitations. Most early research comes with significant limitations like small sample sizes or large cofounders. You have to understand these in conjunction with the results to know how to interpret it.

There’s a cynical approach where people see discussion of limitations, don’t understand it, and instead go into a mode where they think it’s smarter to ignore all criticisms equally because every paper attracts criticisms.

This is just lazy cynicism, though. There are different degrees of criticisms and you have to be able to see the difference between something like a slightly underpowered study, and something like this paper where the authors threw a lot of regressions at a lot of numbers and kind of sort of claimed to have found a trend.

[TheOtherHobbes]

In this case, it only takes a few seconds to find multiple studies confirming the effect.

For example

https://onlinelibrary.wiley.com/doi/10.1111/ina.12042

https://onlinelibrary.wiley.com/doi/10.1111/j.1600-0668.2010...

[conformist]

Conditional on “the study being published and getting attention” the real effect is likely smaller and not larger.

Eg if you assume there is a real effect plus a lot of noise, given the study has been published etc the noise will have more likely acted in the favourable direction.

IMHO given the relatively large size of the effect it seems quite likely that the noise part is in fact potentially large (this is much more subjective) which makes is less clear that there is measurable signal at all here. I’d have to see a lot of replication or a very strong explanation of the underlying mechanism to believe the magnitude of the effect, but will very easily believe the sign (with a small magnitude).

[graemep]

> All research is met on HN by people who know better and will tell you why it's flawed

It is almost certainly flawed, and it is probably wrong: https://journals.plos.org/plosmedicine/article?id=10.1371/jo...

If you are discussing research at all it is important to discuss the flaws too. The alternative I can see would be to take every published paper as proven true even though we know this is not the case.

[bawolff]

Science succeeds when people lean towards the side of cynicism instead of optimism. Scientific research should be read critically.

[mmooss]

Critical thinking and skepticism are good, but much of what happens on HN is not that.

Thinking critically includes, most of all, finding value - you need to think critically (and skeptically) to avoid assigning value to things that don't have it, but you must find value. The goal is to build knowledge - just like the study author needs to find knowledge among flawed data, you must find knowledge among flawed studies - and they are all flawed, of course.

Focusing on the flaws and trying to shoot down everything is just craven recreation.

[jval43]

Science is a long game, it's not about sales where you need to sell right now. Extreme results will be attempted to be replicated, which in turn costs a lot of funding. That is money and time, sometimes a whole persons career.

This money and time is taken directly away from funding other, potentially more worthy or more likely to be correct studies.

There is no point of looking at every (flawed) study in the most positive way, unless you have unlimited time and money to pursue every avenue of research.

Often (not always), the studies that are most heavily promoted among the news and in business or politics are really not the best research and other, less visible but more solid research gets ignored in favor of whats popular or what has had good marketing.

This is very frustrating for people doing solid good research, because every so often someone else will come along with wild, exaggerated claims and very little data to back it up, and then gets funding for it.

It takes literal years away from good science just because someone markets and speaks well.

Which is fine in business, but in science this is not something "the market" can or will correct for well, simply because the timespans are so long.

[mmooss]

> There is no point of looking at every (flawed) study in the most positive way

This line epitomizes the nonsense in the discussion. I didn't say every study, you can't know it's flawed without seriously examining it, and I didn't say in the most positive way at all.

By using these exaggerations, you damage any serious discussion - you give people nothing to respond to except your emotional state.

What I said was, the point is to build knowledge, and so the way to examine research is to find the valuable knowledge - which includes evaluating the accuracy, etc. of that knowledge. There's no other point to it - we're not awarding tenure here, so there's point in keeping some overall score. We just want to learn what we can.

[jval43]

I did not say this study is flawed or that every study is flawed. And I have made no exaggerations or said that you personally look at it the most positive way.

Reading comprehension is important, and especially important in a discussion like this.

I do however really mean that some studies are not worth looking at all in more detail: if the methodology is flawed, the results are meaningless. At most the premise of such a hypothetical (not saying this one necessarily!) study could be used as an idea for further research, but not to build knowledge on or derive knowledge from the results.

[ryandrake]

Are there some examples of "non flawed" research that is getting ignored? Because (as a non-academic) I feel like I'm seeing the same HN attitude that OP describes. No study is good enough for HN. There are always nit pickers that come out of the woodwork. For every science article about some study or finding, the top comment is always a variation on: "This study is flawed because..." Almost without exception. Also, the standard is so high: A single flaw found is grounds for dismissing the whole study as flawed.

My guess is if you raise examples of "good science" the HN peanut gallery will jump in to point out the flaws in that science, too.

[Aurornis]

> you need to think critically (and skeptically) to avoid assigning value to things that don't have it, but you must find value.

This isn’t critical thinking.

This is toxic positivity.

It’s okay to admit that some studies don’t have value to add. If you don’t accept this, you’re going to be tricked by a lot of people trying to get your attention with bad data.

Being able (and willing!) to filter out bad sources, even when they say something you want to hear, is a critically important skill. If you force yourself and others to find something positive about everything then you’re a dream come true to purveyors of low quality or even deliberate misinfo.

[mmooss]

lol

> some studies

It's almost every study on HN, not some studies, which you'd understand if you read my comment.

[jval43]

Yes because most studies that end on up HN are there because they were reported on somewhere as news.

This usually happens in usually these cases:

1. when a paper is extremely good and it's results are groundbreaking, or

2. when a study itself claims it has groundbreaking results, or

3. when it's a regular study that's gotten some great marketing/promotion e.g. by their university.

The case of 1. is extremely rare, and even when everyone believed the results and they were peer reviewed by a reputable paper like Science, some of them turned out to be academic fraud that was later retracted.

Most studies that pop up on HN are of types 2. and 3. That's just because otherwise they would not get news attention.

But most studies in general are in category 4: the ones an academic or professional would read going about their daily business / research. These range from terrible, to OK, to really great, but 99% never make the news.

As a (former) academic, I've read lots of papers and like in real life it's usually the people (papers) that get attention who scream the loudest. There are some gems too of course, and it's right to not ignore anything.

But in my personal experience and over time, I've been very right to be very sceptical once a result turns up in the news because of the 3 ways it can get there.

This is amplified even more so with papers that base their results / outcome purely on statistics, such as most experimental studies done. These derive their results from the statistics (sample size, experiment design, etc) so their power and the probability of their result being correct (what the authors say) it directly coupled.

[strken]

Focusing on the flaws is vital context for helping ordinary people understand the world. Any given study with a surprising result is likely wrong. Yeah, some of them are going to be right, but you're going to get many false positives and only a handful of studies that replicate to a convincing degree.

I've made this mistake time and time again, most recently with vitamin D association studies, and I'm grateful to all the people who urged everyone else to take a wait-and-see approach.

[Jensson]

> Focusing on the flaws and trying to shoot down everything is just craven recreation.

No, its a valuable job to find flaws because its much easier to fix and work on known flaws than to stumble in the dark.

Removing flaws and problems is one of the easiest ways to add value.

[mmooss]

It's not valuable. People who do this at work are people who have no value to offer so they try to sound smart (and valuable) by finding flaws in someone else's work. All work is limited and flawed - it's easy to find them. Add the common hyperbolic statements on HN dismissing the entire study or whole fields of research, and it's misinformation.

The real significance is that things like sample size, to pick a common example here, is easy to understand in a theoretical way and so people apply it to the actual (not theoretical) practice of real research, which they don't understand the practicalities of, and also they overemphasize it because that's pretty much all they understand.

The first thing they look at in a paper is sample size - and hey, now sometimes they have something to 'contribute'! It's just reinforcing the same misunderstandings in others.

It sucks, a little, to have nothing to contribute, but it's a great opportunity to learn from people who do know.

[robertlagrant]

> All research is met on HN by people who know better and will tell you why it's flawed. There isn't a greater collection of expertise in the history of the world than on HN.

It seemed like a pretty valid criticism. These studies should be taken with a massive pinch of salt because they're fairly uncontrolled.

[im3w1l]

Installing air filters is an intervention that has a cost and thus needs to be verified and proven. You don't roll such a thing out on a broad scale based on "the effect could be large".

[dzhiurgis]

I find it bonkers we have better regulation around growing battery chickens than growing kids.

You don’t need massive study to find out that kids don’t like suffocating in classrooms.

It’s a bit like mandating reversing cameras on cars. Study says economically they do not make sense, but not squishing your kids trumps that.

[im3w1l]

Suffocating would be CO2 rather than particles?

[LorenPechtel]

Things which reach the level of getting on here are basically always outliers. And is the outlier real or a false positive? There's a huge selection bias towards the false positives.

[eru]

> Based on your comment, the effect could be larger as well as smaller.

Yes, but since we know that there's a huge bias to publish and publicise larger results, you know what way to bet.

[noosphr]

https://xkcd.com/882/

[lazyeye]

Yes most responses on HN appear to fall into 2 broad categories:-

1. Why this blog post study is flawed and dumb

2. America is bad and shameful.

[mmooss]

That's an interesting comparison, how people analyze and evaluate research and how they do the same with public affairs.

I would have said that the point of research is to find the value and build knowledge, while the point of discussing public affairs is to identify problems to fix.

Thinking about it, I'm not sure the latter can't find things that are constructive. But in either field, the exaggerated, dismissive comments/rants are not just a waste but damaging to progress.

[MPSFounder]

I wish we made more efforts to check air quality. I went to a great school, and the water filter hadnt been replaced since 2005 (it was 2012 at the time, and filters were supposed to be replaced every 6 months). My point is people take it for granted that air quality or water quality in the US is great, but if we checked, I bet most of us are exposed to cancerous materials all the time that could be easily prevented (the water stations rn are made of plastic (PVC), which leaches in the water over time and leads to nanoplastics accumulating in all of us. Easily prevented through more expensive steel pipes). Hell, my office had asbestos in the ceiling that was painted over! It sucks we lack regulation, as I do not believe it is my job to check these things and report them, and I got tired of it. Given the EPA is lacking funds, it is imperative all of us do due diligence so our children do not experience the cancer rates that are ravaging us (our cancer rates are 30% higher than Europe once you rule out lung cancer, which is a result of their smoking habits. The europeans love their cigs).

[eru]

They would have probably been better off without a water filter?

Your local utility should send you water that doesn't need further filtering.

[LorenPechtel]

Steel pipes? You realize steel that can be made into a pipe contains some metals you want to minimize your exposure to?

As for painting over asbestos--that's actually considered acceptable. Asbestos sitting there isn't going to hurt you. Asbestos only hurts you when it's disturbed. Removing intact asbestos is likely to increase your exposure, not decrease it.

As for cancer--why are you so sure it's chemicals and not lifestyle?

[MPSFounder]

I cannot be sure. Causation is very difficult to pinpoint. Not all steel leaches (hell, we use stainless steel cups often). I will take steel over plastics any given day of the week. Regarding paint, sure, but paint peels, and next you are breathing microdoses of asbestos.

[d--b]

Maybe or maybe not. Further research should tell.

Maybe the tests results are better because the children are more rested on the day of the test. Maybe the hum of the machine creates some kind of meditative noise that helps children concentrate. Or maybe none of that is true..

[7thaccount]

I'm a big fan of air filters and have many in my own home that have made a big difference in quality of life as I live in a high pollen area. They can help with a lot more irritants as well that some students may be sensitive to (i.e., some students may study better if their immune system isn't in overdrive for half the school year like mine was). I'm not sure how these would help with natural gas though. I can't read the article due to paywall. Some VOCs can be filtered out (at least I think) with a baking soda filter ...those have to be changed more often than the HEPA filters (at least on my model that has one). Again, that should help with some scents (a major issue for me - even the dishwasher running can cause problems for me), but it isn't going to help if there is a gas leak (not sure if that is what the article is suggesting).

[pizzly]

I'm not sure what it was but I remember learning where with a continuous runny noise in certain classes (buildings) with AC. Outside and different buildings were fine at least for me. In those effected buildings I estimate that 5% were also effected by hearing people sniff every minute or so. This definitely caused distraction while learning. A good filter on the AC would properly have solved this.

[maxerickson]

The article and study are both explicit that natural gas wouldn't have been an issue (the leak was fixed and the gas was gone). The impact from the filters would be from other indoor air pollutants.

[7thaccount]

Thank you!

[LorenPechtel]

The gas was long gone, it wasn't being filtered.

However, I don't see that it proves pollution is the cause. What about infection? Air filtration can reduce the spread of pathogens. Schools throw together a large pool of people, the bug of the day will go around. Less if there are good filters.

[evil-olive]

> but I also think the extreme results from this study aren't going to hold up to further research.

there already is further research. and the results do seem to be holding up.

the study you're quoting from is the one linked in the 2nd paragraph of the article. this is from the 3rd paragraph:

> But it’s consistent with a growing literature on the cognitive impact of air pollution, which finds that everyone from chess players to baseball umpires to workers in a pear-packing factory suffer deteriorations in performance when the air is more polluted.

that paragraph links to an earlier Vox article [0] which goes into more detail, and well as linking to all of the various studies:

> A wide range of studies about the impact of pollution on cognitive functioning have been published in recent years, showing impacts across a strikingly wide range of endeavors. Stripe CEO Patrick Collison has taken an interest in this subject and compiled much of the key research on his personal blog. Among the findings he’s highlighted include:

> - Exposure to fine particulates over the long term leads to increased incidences of dementia and Alzheimer’s disease in the elderly (a second study confirms this).

> - A study of 20,000 older women found that 10 micrograms of additional long-term particulate exposure is equivalent, across the board, to about two additional years of aging.

> - The impacts are not limited to the elderly, however, nor are they exclusively long-term. A range of specialized professionals also seem to suffer short-term impairment due to air pollution. Skilled chess players, for example, make more mistakes on more polluted days. Baseball umpires are also more likely to make erroneous calls on days with poor air quality. Politicians’ statements become less verbally complex on high-pollution days, too.

> - Ordinary office workers also exhibit these impacts, showing higher scores on cognitive tests when working in low-pollution ( or “green”) office environments. Individual stock traders become less productive on high-pollution days.

> - The same also appears to be true for blue collar work. A study of a pear-packing factory found that higher levels of outdoor particulate pollution “leads to a statistically and economically significant decrease in packing speeds inside the factory, with effects arising at levels well below current air quality standards.”

> - Last but by no means least, the cognitive impacts appear to be present in children, with a Georgia study that looked at retrofits of school buses showing large increases in English test scores and smaller ones in math driven by reduced exposure to diesel emissions.

0: https://www.vox.com/future-perfect/2019/12/11/20996968/air-p...

[Aurornis]

> there already is further research. and the results do seem to be holding up.

You’re conflating different results.

The results of the headline study are dramatic in ways that aren’t holding up. The test score increase happened after a single year of putting air filters in class rooms. That’s minimal exposure to purified air for a fraction of the day, 5 days a week, for less than a year of classes.

The other studies are much longe term and look at things like decades of exposure at city scale.

Do you see the difference? The study tried to claim that purifying air immediately improved test scores by dramatic amounts.

There are other serious limitations in the study, like the fact that they can’t even identify which air purifiers were installed or how effective they were. There’s a footnote that says many weren’t even used. There’s a section on air quality monitoring that says they didn’t even detect the VOCs they were trying to filter before they started filtering.

This is the type of study that people implicitly believe because it makes logical sense, but when you read the details you realize that there isn’t much substance in it.

[fc417fc802]

> The other studies are much longe term and look at things like decades of exposure at city scale.

Are they? Several towards the end of the list are specifically about short term day-to-day effects across a range of situations.

> There’s a footnote that says many weren’t even used.

Quite damning if true.

> they didn’t even detect the VOCs they were trying to filter before they started filtering

That is confusing the reasoning that led to the company footing the bill versus what was being looked at here. The lack of VOCs is actually in the author's favor as it eliminates "exposure to atypical VOCs" as an otherwise fairly severe confounding factor.

[xattt]

If anything, schools able to implement air filtration and fresh air exchanges systems are likely those flush with cash and supportive parents.

[permo-w]

in this case the filtration systems were installed on the dime of a local natural gas company after a leak, although that was as a result of the interference of interested parents

[dzhiurgis]

These systems are not that expensive and save money.

If parents would be allowed to contribute towards it like a 10year bond it would pay for itself…

[xwolfi]

And the resulting backlash 10 years down the line ? I don't think science should be used that way: who would trust it eventually ?

Would you support a small study saying a medicine or a vaccine produces a 20 year life expectancy increase, all that to end up 20 years later with no improvement, everyone on that medicine, and the anti-everything yelling on every platform that the big pharma lobby poisoned our children ?

Even when the studies are on large samples, double-blind, long time range with a clear explanation as to why there's an effect, we have people trying to kill the resulting health campaigns. Don't encourage fake ones !

[eru]

> I don't think science should be used that way: who would trust it eventually?

There's no single party controlling 'science'. It's all just individuals many of them under 'publish or perish' rules.

[calvinmorrison]

only ycombinator would have a hard time with the concept of 'open the window' and instead expect a capital installment of $2B to retrofit classrooms

[do_not_redeem]

Do you think everyone lives in California where there's comfortable sunny temperatures and no rain 360 days a year? Classic ycombinator post

[eru]

German schools open their windows in winter.

It's the classic German Stoßlüften in action.

However, the air outside also has lots of dusts. Opening the windows does nothing for that, but air filtration can potentially help.

[dzhiurgis]

We did it too, but it was huge ceremony around it. Kids shouldn’t be begging and screaming “too cold” for fresh air.

[eru]

The ceremony is part of the fun.

Though for us the ceremony was only one or two mild complaints from the kids near the window, and the teacher firmly stating that the room reeks of sweaty teenagers and needs some airing out.

[nottorp]

When I was in school we opened the windows to let "fresh air" in even at -20 C... not all day ofc but every day.

And rain? What's the problem with opening a window when it's raining? Only if there's a rain storm... and not even then, depending on wind direction.

[lazyasciiart]

Yes, silly gas company, should just tell the local schools to open the windows and let the natural gas in.

You have heard of air pollution, I assume. Some schools are in places that regularly have polluted air.

[ajb]

In many areas the source of air pollution is outside. Opening windows helps with house dust and mold spores, but not with particulates from diesel, tyres or with pollen.

[dzhiurgis]

Perhaps we shouldn’t need a large cohort study to give little humans the right to breathe as good as battery chickens.

[ninetyninenine]

You ever heard of the paradox of IQ?

IQ has about a 0.2 correlation with income. The paradox arises when you zoom out for a more macro view. National IQ has about a 0.6-0.8 correlation with GDP per capita.

Performance in class rooms is definitely an IQ thing and different view points will likely generate different sets of data.

I'm interested in seeing controlled trials on individual performance, not just observing real world scenarios.

[VladVladikoff]

One problem with nutrition based science is it suffers from nutrigenomics bias. It’s possible to study small cohorts of similar genetics and come to conclusions which fail to extrapolate to larger populations. It’s possible that this same problem does not apply to air filtration.

[ted_dunning]

It's possible, but much less likely since they controlled for demographics and cited a number of related studies that show comparable effects.

[fasthands9]

It also seems possible that students do better on tests when the air is cleaner, but that doesn't necessarily mean the students learned more.

Imagine if some schools installed air conditioning in their gym one year. Running times around an indoor track would improve considerably, but mostly because conditions at the point of testing improved. Not necessarily because the air conditioning made the students actually improve their stamina or speed.

[facile3232]

> but expecting test scores to improve dramatically like this is not realistic

That seems like a problem for the reader, not a problem with the text. Why would the reader expect this? Is it the use of present tense in the title rather than past tense?

[Aurornis]

What do you mean? The article implies it from the title through the text.

They hedge by saying “could” but like most articles in this vein it goes on to pivot around the outlier study.

[facile3232]

Ah! I just skipped the article. The paper is much more straightforward.