Very hard to see how they can discount confounding correlates to diesel particulate and other PM10 and finer materials, brake dust &c.
Not that I don't think the science is good, I'm sure they do and the article says they do, but I struggle to see how you can isolate noise from particulate exposure: they'd have semi identical square law dropoff by distance and volume.
1. Electric trains wont generate diesel particulate, but will generate road noise.
They don't generate diesel particulate, by they do generate "other PM10 and finer particulate", from the brakes for example.
In Paris there's a lawsuit against the RATP (local transit authority) for particulate pollution inside the stations. [0]
There's also the fact that, at least in Paris, some metros are running on tires. For those that run on metal wheels and rails, some turns are quite sharp, and you can hear the screeching of the metal. This is likely to contribute to particulate pollution.
I suppose for overground trains, this is less of an issue.
Except road noise was associated with dementia and train noise wasn't, which suggests it's pollution related according to your (1)
The other problem is that people that live near noise are poorer. Did they control for wealth?
These epidemiological observational studies are so hard to take seriously. I get why so many of them are done though. It's a relatively lower effort way to get publications out the door, all you need is a dataset and a few days in python.
Of course they controlled for wealth (individual income, occupational status, highest attained education and neighbourhood socioeconomic status).
It's hard to take seriously HN commenters who assume people working outside software development are all idiots who need to be told how to do their job.
Epidemiological studies in general are flawed and hard to trust due to omitted variables that you don't know about, p-hacking and publication bias making it difficult or impossible to interpret causality. You say "of course" as if it's a foregone conclusion that the results should be trusted and the authors have done a good job, when the default should be disbelief and scepticism in this particular approach unless shown otherwise with unusually strong results. I only have a positive reaction and belief by default in large scale RCTs. Epidemiological studies have not earned the benefit of the doubt. At best they are indications of future experimental research directions, i.e. they are a means to an end.
I've worked in academic applied statistics research and seen how the cookie crumbles, it's not pretty, and yes there are groups and industries of people who may be individually intelligent but nevertheless act like idiots given the incentives they face to churn out low-effort publications. Search for "hegemony" or "Fuzzy neural net Dow Jones" in Google Scholar to see 140 IQ people mass produce drivel. Just because I am not trained in medicine doesn't mean I am not allowed to draw this conclusion, I have years of experience in statistics and know a bullshit application of these tools when I see it, and I don't appreciate appeals to authority or other arguments that try to invalidly shut down people's opinions without knowing why those opinions were formed in the first place.
“The results are strengthened by the authors controlling for levels of traffic pollution, which was recently shown to be associated with dementia as well as other diseases.”
Skimming the article [0], it looks like they compared outcomes of people who lived closer to the road against people who lived farther away from the road, in the same building.
I think the idea is that people living in the same building should all have roughly the same particulate and pollution exposure, but people living on a road-facing side of the building will have more exposure to road noise. It's an extremely clever study design.
If the assumption is that people in the same building have the same exposure to pollution but differential exposure to noise based on distance, I would prefer that it were tested. I am sceptical that this is the case, as I would expect folk on the road facing side to be exposed to more particulates and pollution and noise.
That's a good point, and I didn't read far enough in the paper to see if they tested that assumption or not. It seems like they developed some kind of model of noise exposure as a function of elevation and proximity to the road, so it's possible they developed a similar model for pollution.
Comments
Very hard to see how they can discount confounding correlates to diesel particulate and other PM10 and finer materials, brake dust &c.
Not that I don't think the science is good, I'm sure they do and the article says they do, but I struggle to see how you can isolate noise from particulate exposure: they'd have semi identical square law dropoff by distance and volume.
Two ways I can think of:
1. Electric trains wont generate diesel particulate, but will generate road noise.
2. It's relatively straightforward to measure pollution and noise, so you can do analysis to try and separate them.
In many situations these would not be correlated exactly. For example, terrain, wind patterns, noise dampening structures, etc.
They don't generate diesel particulate, by they do generate "other PM10 and finer particulate", from the brakes for example.
In Paris there's a lawsuit against the RATP (local transit authority) for particulate pollution inside the stations. [0]
There's also the fact that, at least in Paris, some metros are running on tires. For those that run on metal wheels and rails, some turns are quite sharp, and you can hear the screeching of the metal. This is likely to contribute to particulate pollution.
I suppose for overground trains, this is less of an issue.
---
[0] https://www.rfi.fr/en/france/20210323-french-ngo-sues-paris-...
They could adapt the system of
[1] https://en.wikipedia.org/wiki/Montreal_Metro#Design
Popcorn!
Except road noise was associated with dementia and train noise wasn't, which suggests it's pollution related according to your (1)
The other problem is that people that live near noise are poorer. Did they control for wealth?
These epidemiological observational studies are so hard to take seriously. I get why so many of them are done though. It's a relatively lower effort way to get publications out the door, all you need is a dataset and a few days in python.
Of course they controlled for wealth (individual income, occupational status, highest attained education and neighbourhood socioeconomic status).
It's hard to take seriously HN commenters who assume people working outside software development are all idiots who need to be told how to do their job.
Epidemiological studies in general are flawed and hard to trust due to omitted variables that you don't know about, p-hacking and publication bias making it difficult or impossible to interpret causality. You say "of course" as if it's a foregone conclusion that the results should be trusted and the authors have done a good job, when the default should be disbelief and scepticism in this particular approach unless shown otherwise with unusually strong results. I only have a positive reaction and belief by default in large scale RCTs. Epidemiological studies have not earned the benefit of the doubt. At best they are indications of future experimental research directions, i.e. they are a means to an end.
I've worked in academic applied statistics research and seen how the cookie crumbles, it's not pretty, and yes there are groups and industries of people who may be individually intelligent but nevertheless act like idiots given the incentives they face to churn out low-effort publications. Search for "hegemony" or "Fuzzy neural net Dow Jones" in Google Scholar to see 140 IQ people mass produce drivel. Just because I am not trained in medicine doesn't mean I am not allowed to draw this conclusion, I have years of experience in statistics and know a bullshit application of these tools when I see it, and I don't appreciate appeals to authority or other arguments that try to invalidly shut down people's opinions without knowing why those opinions were formed in the first place.
The article says:
“The results are strengthened by the authors controlling for levels of traffic pollution, which was recently shown to be associated with dementia as well as other diseases.”
How can you possibly control for traffic pollution in a way that is independent from traffic noise?
Skimming the article [0], it looks like they compared outcomes of people who lived closer to the road against people who lived farther away from the road, in the same building.
I think the idea is that people living in the same building should all have roughly the same particulate and pollution exposure, but people living on a road-facing side of the building will have more exposure to road noise. It's an extremely clever study design.
0: https://www.bmj.com/content/374/bmj.n1954
If the assumption is that people in the same building have the same exposure to pollution but differential exposure to noise based on distance, I would prefer that it were tested. I am sceptical that this is the case, as I would expect folk on the road facing side to be exposed to more particulates and pollution and noise.
That's a good point, and I didn't read far enough in the paper to see if they tested that assumption or not. It seems like they developed some kind of model of noise exposure as a function of elevation and proximity to the road, so it's possible they developed a similar model for pollution.
With ear plugs apparently. But then you have an entirely different control problem.
We dont have an accurate assessment of traffic pollution or air pollution in general. Thanks 'regulatory agencies'