I think that a reddit/hn-style voting system is the absolute opposite of what you want to have for scientific publications. These systems tend to enforce a strong hive-mind, where only the most popular posts and opinions are visible and anything else gets downvoted into oblivion.
don't citations already function like a voting system (albeit with no downvotes)? a paper with zero citations will be ignored just as surely as a comment in a large thread with zero votes.
Because democratic research is certainly going to do wonders in an era where the equipment and/or resources to reproduce any of the claims of these papers is way inaccessible to the average voter.
I'd put my finger that even random sorting would bring more value.
It's not exactly that, but with Plaudit [1] you're able to see which academics have read it and considered it worthwhile enough to endorse. Ideally bioRxiv/medRxiv would integrate it natively like several other preprint servers have done (e.g. [2]), but for now you'll need an extension though.
How would that work? Would random people be able to vote on which research papers have the most merit? Only people with relevant doctorates?
Another commenter mentioned a projected call Plaudit, which seems to use the 'ORCiD' system to decide who is worth taking seriously. [0] Can anyone explain how that really works?
Ask HN: Has any study been done doing a random sampling of a population using an antibody immunity test? It seems the natural next step given we have such a test.
Until we do that, I feel like we have no idea of the actual infection rate and population immunity.
We know for a fact that ICUs are being overloaded with cases and morgues are overwhelmed with the dead. Arguing over exact figures is like quibbling over your exact velocity as you're falling off a cliff.
Finding some prettier numbers won't change the reality of the situation, no matter how convincingly you argue your case.
People are talking doomsday scenarios, on this very forum. Getting some hard facts would really help temper the hysteria, or allow people to prep further appropriately.
We’re planning for more than the next three days here.
An alternative to your alternative: science and medicine still work. Let's try and make it better. The better we understand our situation the better we can react appropriately.
Yes, it's really bad in a few places right now. It looks like it's going to be really bad in a lot more places in the next 15-20 days. How _long_ will it be really bad for and for how _many_ people will it be really bad are important questions to understand.
I'm not being irrational; you are. I accept what I'm seeing as the reality of the situation. You appear to be looking for numbers that will tell you things are different to the observed reality. That's my point.
The situation in South Korea is clearly an outlier. Even knowing exactly the steps which have drastically slowed the growth of the virus in South Korea is about as useful as knowing last week's lottery numbers at this point.
It's not an outlier. Germany has even lower numbers despite doing even less than Korea.
Italy and the UK book the death of anyone who died testing positive as a covid19 death. The UK won't release any figures beyond that. Not even hospitalisation data. Studies in Italy show fewer than 15% of those deaths might be specifically caused by it.
> On re-evaluation by the National Institute of Health, only 12 per cent of death certificates have shown a direct causality from coronavirus, while 88 per cent of patients who have died have at least one pre-morbidity - many had two or three,” he says.^
In fact the average comorbiditities is 2.7. mean age of death is 79.5 years. Median 80.5.
If many more people are infected with mild symptoms than currently thought, that could have large implications for how long economic shutdowns need to last.
Well it's currently thought as many as 50% don't develop symptoms for the duration of the infection (and likely aren't necessarily a big infection risk) based on studies in Italy. 90% only very mild symptoms.
Yes, that's what I think. The actions two weeks ago were:
Shut down everything except essential services, isolate everyone as much as possible, throw as many resources as possible at your health service. Even then, they will be overwhelmed at some point.
Which country are you even talking about? If you think the US is literally doing everything they can, have you looked at what the lockdown in Italy looks like?
If you have random sampling from two weeks ago and today, you might find out that you won't have to go to the same extremes as italy, or you might find out that military should have been on the streets a week ago and putting them on the streets now is a good idea
I'm talking generally about all developed countries (I know what in many developing countries a lockdown will simply be impossible).
I certainly don't think the US has done everything it can. Quite the contrary - it's no surprise to me at all that it is now top of the leaderboard in recorded cases. The US had to go to at least the extremes of Italy, and it didn't, so you'll end up with many more deaths than you would otherwise have had.
The US could just have looked at what was actually happening in Europe to establish what it needed to do. Arguing over the numbers is exactly what caused the inaction in the first place.
Ok, then let's talk Europe. Denmark has been in a early and disciplined lockdown for 3 weeks. The amount of deaths per day seem to have stabilized. The Hospitals are not overrun
When should Denmark end the lockdown?
(And if Denmark had a random sampling from two weeks ago, and one from today, would they be better positioned to know when to end the lockdown or not?)
No, that's completely wrong. Denmark only had its first death 2 weeks ago (a single death) and one the following day. Over the last 3 days, it's had 31 deaths. The number of active cases is still shooting up. It's very early days for Denmark and things will get a lot worse before it gets better. I've no idea why you think anything has stabilised there.
Is it perhaps the change in daily new cases that's thrown you? Many countries have that drop and increase again as they change the way they're testing.
... Sigh. Ok then the cases in Denmark hasn't stabilized. How in the world does that change my point?
(but number of daily deaths have been linear for a week now, I don't know why you'd think that's not stable, but you're just completely proving my point by disagreeing with how stable the situation is in Denmark, how do you ACTUALLY know that it's not?)
Number of daily deaths for the last 3 days, as I mentioned before, was 31. The 3 days before that, 18. The 3 days before that 15. That is clearly increasing.
I know it's not from the statistics. Note I don't need to know the exact percentage of asymptomatic cases for this, or to know the correct course of action. Which was my point to start with. Perhaps you lost track of it?
So, what is it going to be in 3 days? 6 days? 9 days? Your view on the situation is purely reactionary. Having population-level data allows public health officials to proactively take steps to mitigate the current crisis and anticipate how much surge capacity they will actually need. Right now they can't do that, because even testing of symptomatic people is bottlenecked.
Being reactionary to a crisis like this is a very bad idea and will lead to it being worse than it needs to be.
I think that fear that people are on fishing expedition to find any numbers that could be twisted as optimistic and then use them for politics is not reactionary paranoia. That is exactly what was and is going on last weeks.
And phaemon is completely correct in saying that US has not done everything it can. It was the opposite - spinning situation in absurdly optimistic way until it became impossible to ignore.
Everybody should care. The most important thing, knowing the situation would enormously help with shaping up the response. If 0.0005% of the population had COVID-19 would require different actions compared to a situation where 30% had it.
Not everywhere ICUS are being overloaded and morgues are overwhelmed with the dead. And ignorance was the first reason why it happened where that's the case. So anything that could decrease the ignorance should be met with open arms.
That's exactly what I mean. There is no scenario where that is the case. If your numbers tell you that, your numbers are wrong.
The actions are the same regardless: shut down everything except essential services, isolate everyone as much as possible, throw as many resources as possible at your health service. Even then, they will be overwhelmed at some point.
What numbers exactly would lead you to a different course of action, based on what we have already observed?
Because if the number was 0.0005% you wouldn't see ICUs and morgues overwhelmed (why don't you calculate how many people that is for your country?)
And, again, that's my point: you're asking why it can't be a ridiculously small number when it should be obvious from the reality of the situation that it isn't!
Sorry, you make zero sense. First of all, not everywhere are ICUS and morgues overwhelmed. Some towns in Italy and Spain are hit bad, South Korea manages it without even having lockdowns.
Second, and that's my point: you can have the same result with two very different scenarios: either a lot of people have/had it but in the vast majority of the cases it goes without symptoms at all. So when a big part of the population gets infected only then you start noticing "ICUs being overwhelmed".
The second scenario is: it causes serious problems for a substantial part of the infected so even the small number of those who got the virus will flood the hospitals.
And those two scenarios require different responses.
And the percentage of really asymptotic cases is much lower than the most here expect.
All the tests in all the countries could be seen as the statistical sampling. The speed of the transmission is also known, and surely not fundamentally different than the speed of people needing hospital or dying -- the doubling time is nowhere faster than 2 days, and the typical is 3 days.
In short, it can't be expected that "already everybody has it" in the positive sense of "nothing will happen."
Because it makes a hell of a lot of difference if we are seeing those numbers in the ICU when 5% of the population is infected or 50% is infected. One means the situation is going to get way way worse in the future and one means we're probably nearing the peak now.
Actually, I've changed my mind. I still don't think an antibody test should be a prerequisite for current action, but I think it will be very useful for deciding what to do in a few weeks.
That's the point I was trying to make (quite clumsily, I agree). I hope I didn't give the impression that I expected the numbers to prove that we are overreacting. Having a reasonable idea of how many people have been infected will help when trying to avoid a second wave of infections
3 blue one brown had a great video on exponential growth that detailed how a 10% difference in the numbers is the difference between being halfway there and having orders of magnitude to go.
As I'm sure you're aware, the point of the numbers isn't to change the reality of the situation, it's to understand what that reality is. And the people that care are epidemiologists and leaders with decades of experience fighting pandemic after pandemic. You know, people that know what they're talking about? Might be worth trying to spend time actually learning how the people that are spending day and night trying to save your life and the lives of your loved ones do their work before dismissing them as useless.
I didn't. Try reading what I wrote. I'm dismissing half-brained idiots who try to downplay the seriousness of the situation by playing with numbers while ignoring what those numbers mean.
Probably the false-positive rate of antibody tests would be a big problem as long as only a small fraction of the population is infected. From what I've read, the specificity of such tests is way worse than for RNA-based tests.
Last I checked they had a roughly 10% false positive rate. So if 1 in 1000 people were infected but you tested all 1000 with an antibody assay, you would get 100 positives, and probably one of them is actually infected. Basically useless for mass screening. The gold standard RT-PCR assay has basically zero false positives, but requires experienced personnel and expensive reagents to do en-mass.
The RTPCR will only work for those with viral RNA in them. So those with active infections. The antibody test won't work until the host produces antibodies against the virus (except for false positives). So yeah, the antibody test will detect those who are immune, or at least have antibodies against Covid-19 whereas the RT-PCR will not.
I keep seeing news articles along the lines of "People getting reinfected" "In 5% of cases recovered patients tested positive for COVID-10 2 months later" "Asymptomatic individuals test positive for COVID anyway" Yeah very mysterious indeed :)
Why would you think that people couldn't get reinfected ?
The virus won't just bounce off your skin after one has got into you. If it is present everywhere in the environment then you will keep getting infected, maybe faster than your immune system can deal with it.
The end point to aim for in terms of immunity is that your immune system can deal with any level of viral load in the future without you becoming ill. Maybe combined with prophylactic drugs that help prevent the virus reproducing as fast within you.
>100% Konformitatsrate (Spezifitatsrate) fur 126 Negativkontrollen mit nicht mit SARS-CoV-2 akut oder vorher infizierten Probanden.
Heh, well that's great news. It's an antibody test so it would tell you who has had antibodies against the virus and thus who is probably immune to it.
They're planning to do exactly that in some (probably many) countries. But the antibody tests are just being developed and might be ready in the next weeks for larger scales. But you need to do this properly to get a representative sample, which takes a bit of time.
The Dutch blood bank, Sanquin, is currently running a study now for 10,000 blood donors, to see if the Netherlands is close to achieving 'herd immunity'
This will be important information in a few months when countries have lowered the new detected cases and the lift of harsh measures is considered. So there is still some time to get the data. At the current moment, they are not important for the decisions.
However I want to caution anyone to expect a large proportion of the population to test positive in these tests.
We know that South Korea was able to contain the spread before a significant proportion of the population was infected. A important part of the strategy was contact tracing. For contact tracing to work, you need to detect at least 1-1/R of cases. If we assume that other measures such as masks and limited social distancing pushed R down to 2 and assume they were able to trace 100% of contacts, then we still have only twice as many that are infected. If nobody of the cases that are still ongoing dies, we look at a lower bound of the IFR of 0.8%.
How can you say this? If it's significantly more infectious than we thought, that will obviously influence public policy as the r_0 will be radically decreased
The number of people that are infected doesn't tell you that much about R0. You can get R0 from the increase in daily new deaths and the generation time between infections, given that the quality of care stays constant.
Serological tests will tell you something about the early outbreak and how many cases where not detected and how high the infected fatality ratio really is.
I have seen many people that still believe that the IFR is in the range of an average flu. With the data from South Korea, we can put a number at what we can expect the IFR to be if we make some optimistic assumptions.
That is why I say that you should not expect that serological tests show that a large part of the population already had covid19 and that the IFR is much lower than we expect. They will show where exactly the IFR is.
Or even the Infection Fatality Rate. The only way to stop fumbling in the dark is to get a random representative sample tested. It’s not impossible there is a dark pool of great numbers of present and past infections that are causing people to overreact .
On the days where I was still energetic enough to try and debunk some of the pseudosciences, I quickly discovered that one of the primary rules is absolutely never to trust one single article, however "scientific", and however reputable the publisher of the journal/conference is.
Otherwise one quickly becomes overwhelmed by the amount of crap published by otherwise "respectable" publishers (e.g. Elsevier's Journal of Homeopathy). Let's not even get into non-reviewed articles...
The only things that _may_ have any value whatsoever are the ones published in utmost respectable _journals_ themselves (e.g. Nature, but absolutely not on any of their spinoffs), and/or when there are very big labs/organizations in the authors list. Very big. I mean usually national pharmaceutical associations (obviously not from all states) or huge corporations (who will generally avoid putting their name on the line either way).
Meta-analyses most importantly do NOT have any intrinsinc value either _just because_ they are meta-analyses.
By reading bioXiv directly, you are most likely doing yourself a disservice. Unless you are a state agency or huge corporation having huge resources backing your which allows you to internally replicate whatever they claim.
You're going too far in dismissing the value of individual scientific papers. If you read the methods and analysis yourself you can figure out how good their results are. If you read a lot of papers or read several meta-analyses you can form a opinion about what the consensus is. All of these options are more work than reading one abstract and calling it a day but that doesn't mean scientific papers are worthless.
If scientific papers did not contain any information then scientists would not read them.
No, you cannot figure out by yourself "how good the results are". That is the one of the critical points. The paper might even pass peer review and get accepted -- which, assuming a reasonable quality journal, means that the methods and analysis they claim make sense.
But that is just the _very minimum_ you can expect in an article. Incredibly enough, by itself, that is already a huge filter. But sadly that is nowhere near enough to give any value whatsoever to the article. i.e. the article can be perfect and still just wrong. Even assuming good faith from the authors, they could just be the lucky 5% who got a positive out of an impossible thing and decided to publish. In a world with a general pressure to publish positive results and specially in a highly-hyped topic such as this one, you have a very high chance of seeing such a result.
At the end of the day this is the only argument I have left to debunk some homeopathy papers.
Scientists who are in a position where they have the resources to _reproduce_ the articles should and do read them. "Mortals" are not and reading these articles is a waste of time and potentially a disservice.
There are tens of journals claiming _utter non-sense_ being published every single month.
Using un-reviewed papers to estimate "consensus" is also pointless. Using publication numbers / citations / impact factors is slightly better but also extremely dangerous. If anything pseudo-science chaps are good at is at publishing large numbers of astonishing crap.
The sad truth is that filtering good science vs not is out of reach for the average layman, which leaves you with unscientific stuff such as "reputation" as your only tool.
I think you might want to qualify that by specific fields. For example if the paper you're reading is math or theoretical physics, then you actually do have all of the same capabilities as the author to verify it. As for publication bias, some studies are pre-registered.
>Scientists who are in a position where they have the resources to _reproduce_ the articles should and do read them.
It is typical for a scientist to read hundreds and hundreds of papers every year, or sometimes even every month. Are you suggesting that they reproduce every one of them? You may be dismayed to find out that typically, they would reproduce none of them, and when the system is working its best they might reproduce one of them. Under usual circumstances, you and they have exactly the same replication ability: the ability to search the literature for a replication.
While the general sentiment and advice you give is mostly correct, I feel like adding my two cents from my experience -- as a person who occasionally publishes in these sorts of journals.
In my opinion, you have the Science/Nature thing vs their spinoff exactly the other way around. Science/Nature only publish flashy things more for their "memeness" value than their actual scientific content. The first driver leading you to those journal is how much the paper can be talked about and picked up by other media outlets. In fact, the most important thing you need to publish in Science/Nature is NOT a good experiment/result, but being a coauthor of somebody who already has a lot of Science/Nature papers. Their spinoffs, on the other hand, being less driven by popularity, often offer more solid content.
THanks for these. I am interested in modeling this effort, and it is difficult to find papers that outright call R00, population logistic saturation parameters, case resolution timescale, detection probabilities (ie if you have x tests per million ppl, what is the actual number of cases), etc.
R00 that I typically use is about 2.1-2.6, however relating this to the 30-60% case increase per day is a difficult task.
If you know more information about where I could find these I would appreciate it.
73 comments
[ 4.6 ms ] story [ 133 ms ] threadEdit: And HN-style comments might make it easier to discuss the papers.
I'd put my finger that even random sorting would bring more value.
(Disclosure: I'm a volunteer working on Plaudit.)
[1] https://plaudit.pub/extension/
[2] https://psyarxiv.com/y38m9/
Another commenter mentioned a projected call Plaudit, which seems to use the 'ORCiD' system to decide who is worth taking seriously. [0] Can anyone explain how that really works?
[0] https://plaudit.pub/#concept
Until we do that, I feel like we have no idea of the actual infection rate and population immunity.
We know for a fact that ICUs are being overloaded with cases and morgues are overwhelmed with the dead. Arguing over exact figures is like quibbling over your exact velocity as you're falling off a cliff.
Finding some prettier numbers won't change the reality of the situation, no matter how convincingly you argue your case.
We’re planning for more than the next three days here.
Yes, it's really bad in a few places right now. It looks like it's going to be really bad in a lot more places in the next 15-20 days. How _long_ will it be really bad for and for how _many_ people will it be really bad are important questions to understand.
Don't let the dire situation make you irrational.
Italy and the UK book the death of anyone who died testing positive as a covid19 death. The UK won't release any figures beyond that. Not even hospitalisation data. Studies in Italy show fewer than 15% of those deaths might be specifically caused by it.
> On re-evaluation by the National Institute of Health, only 12 per cent of death certificates have shown a direct causality from coronavirus, while 88 per cent of patients who have died have at least one pre-morbidity - many had two or three,” he says.^
In fact the average comorbiditities is 2.7. mean age of death is 79.5 years. Median 80.5.
^ https://www.telegraph.co.uk/global-health/science-and-diseas...
Shut down everything except essential services, isolate everyone as much as possible, throw as many resources as possible at your health service. Even then, they will be overwhelmed at some point.
The same actions apply today. What's changed?
If you have random sampling from two weeks ago and today, you might find out that you won't have to go to the same extremes as italy, or you might find out that military should have been on the streets a week ago and putting them on the streets now is a good idea
I certainly don't think the US has done everything it can. Quite the contrary - it's no surprise to me at all that it is now top of the leaderboard in recorded cases. The US had to go to at least the extremes of Italy, and it didn't, so you'll end up with many more deaths than you would otherwise have had.
The US could just have looked at what was actually happening in Europe to establish what it needed to do. Arguing over the numbers is exactly what caused the inaction in the first place.
When should Denmark end the lockdown?
(And if Denmark had a random sampling from two weeks ago, and one from today, would they be better positioned to know when to end the lockdown or not?)
Look at the graphs: https://www.worldometers.info/coronavirus/country/denmark/
Is it perhaps the change in daily new cases that's thrown you? Many countries have that drop and increase again as they change the way they're testing.
(but number of daily deaths have been linear for a week now, I don't know why you'd think that's not stable, but you're just completely proving my point by disagreeing with how stable the situation is in Denmark, how do you ACTUALLY know that it's not?)
I know it's not from the statistics. Note I don't need to know the exact percentage of asymptomatic cases for this, or to know the correct course of action. Which was my point to start with. Perhaps you lost track of it?
Being reactionary to a crisis like this is a very bad idea and will lead to it being worse than it needs to be.
And phaemon is completely correct in saying that US has not done everything it can. It was the opposite - spinning situation in absurdly optimistic way until it became impossible to ignore.
Not everywhere ICUS are being overloaded and morgues are overwhelmed with the dead. And ignorance was the first reason why it happened where that's the case. So anything that could decrease the ignorance should be met with open arms.
That's exactly what I mean. There is no scenario where that is the case. If your numbers tell you that, your numbers are wrong.
The actions are the same regardless: shut down everything except essential services, isolate everyone as much as possible, throw as many resources as possible at your health service. Even then, they will be overwhelmed at some point.
What numbers exactly would lead you to a different course of action, based on what we have already observed?
And, again, that's my point: you're asking why it can't be a ridiculously small number when it should be obvious from the reality of the situation that it isn't!
Second, and that's my point: you can have the same result with two very different scenarios: either a lot of people have/had it but in the vast majority of the cases it goes without symptoms at all. So when a big part of the population gets infected only then you start noticing "ICUs being overwhelmed". The second scenario is: it causes serious problems for a substantial part of the infected so even the small number of those who got the virus will flood the hospitals.
And those two scenarios require different responses.
https://www.eurosurveillance.org/content/10.2807/1560-7917.E...
And the percentage of really asymptotic cases is much lower than the most here expect.
All the tests in all the countries could be seen as the statistical sampling. The speed of the transmission is also known, and surely not fundamentally different than the speed of people needing hospital or dying -- the doubling time is nowhere faster than 2 days, and the typical is 3 days.
In short, it can't be expected that "already everybody has it" in the positive sense of "nothing will happen."
So the true numbers are definitely higher than 0.0005%.
As I'm sure you're aware, the point of the numbers isn't to change the reality of the situation, it's to understand what that reality is. And the people that care are epidemiologists and leaders with decades of experience fighting pandemic after pandemic. You know, people that know what they're talking about? Might be worth trying to spend time actually learning how the people that are spending day and night trying to save your life and the lives of your loved ones do their work before dismissing them as useless.
https://youtu.be/Kas0tIxDvrg
I don't need any self-righteous advice from you.
https://projecteve.com/organizing-tech-tools-for-work/#comme...
The virus won't just bounce off your skin after one has got into you. If it is present everywhere in the environment then you will keep getting infected, maybe faster than your immune system can deal with it.
The end point to aim for in terms of immunity is that your immune system can deal with any level of viral load in the future without you becoming ill. Maybe combined with prophylactic drugs that help prevent the virus reproducing as fast within you.
https://www.nature.com/articles/d41587-020-00010-2
https://pharmact.eu/
Don't know if this would show immunity though.
Heh, well that's great news. It's an antibody test so it would tell you who has had antibodies against the virus and thus who is probably immune to it.
I understand the need for a proper study, just surprised we're not seeing more on this.
The original report from Friday, auf Deutsch: https://www.spiegel.de/wissenschaft/medizin/coronavirus-gros...
However I want to caution anyone to expect a large proportion of the population to test positive in these tests.
We know that South Korea was able to contain the spread before a significant proportion of the population was infected. A important part of the strategy was contact tracing. For contact tracing to work, you need to detect at least 1-1/R of cases. If we assume that other measures such as masks and limited social distancing pushed R down to 2 and assume they were able to trace 100% of contacts, then we still have only twice as many that are infected. If nobody of the cases that are still ongoing dies, we look at a lower bound of the IFR of 0.8%.
Serological tests will tell you something about the early outbreak and how many cases where not detected and how high the infected fatality ratio really is.
I have seen many people that still believe that the IFR is in the range of an average flu. With the data from South Korea, we can put a number at what we can expect the IFR to be if we make some optimistic assumptions.
That is why I say that you should not expect that serological tests show that a large part of the population already had covid19 and that the IFR is much lower than we expect. They will show where exactly the IFR is.
On the days where I was still energetic enough to try and debunk some of the pseudosciences, I quickly discovered that one of the primary rules is absolutely never to trust one single article, however "scientific", and however reputable the publisher of the journal/conference is.
Otherwise one quickly becomes overwhelmed by the amount of crap published by otherwise "respectable" publishers (e.g. Elsevier's Journal of Homeopathy). Let's not even get into non-reviewed articles...
The only things that _may_ have any value whatsoever are the ones published in utmost respectable _journals_ themselves (e.g. Nature, but absolutely not on any of their spinoffs), and/or when there are very big labs/organizations in the authors list. Very big. I mean usually national pharmaceutical associations (obviously not from all states) or huge corporations (who will generally avoid putting their name on the line either way).
Meta-analyses most importantly do NOT have any intrinsinc value either _just because_ they are meta-analyses.
By reading bioXiv directly, you are most likely doing yourself a disservice. Unless you are a state agency or huge corporation having huge resources backing your which allows you to internally replicate whatever they claim.
If scientific papers did not contain any information then scientists would not read them.
But that is just the _very minimum_ you can expect in an article. Incredibly enough, by itself, that is already a huge filter. But sadly that is nowhere near enough to give any value whatsoever to the article. i.e. the article can be perfect and still just wrong. Even assuming good faith from the authors, they could just be the lucky 5% who got a positive out of an impossible thing and decided to publish. In a world with a general pressure to publish positive results and specially in a highly-hyped topic such as this one, you have a very high chance of seeing such a result.
At the end of the day this is the only argument I have left to debunk some homeopathy papers.
Scientists who are in a position where they have the resources to _reproduce_ the articles should and do read them. "Mortals" are not and reading these articles is a waste of time and potentially a disservice.
There are tens of journals claiming _utter non-sense_ being published every single month.
Using un-reviewed papers to estimate "consensus" is also pointless. Using publication numbers / citations / impact factors is slightly better but also extremely dangerous. If anything pseudo-science chaps are good at is at publishing large numbers of astonishing crap.
The sad truth is that filtering good science vs not is out of reach for the average layman, which leaves you with unscientific stuff such as "reputation" as your only tool.
>Scientists who are in a position where they have the resources to _reproduce_ the articles should and do read them.
It is typical for a scientist to read hundreds and hundreds of papers every year, or sometimes even every month. Are you suggesting that they reproduce every one of them? You may be dismayed to find out that typically, they would reproduce none of them, and when the system is working its best they might reproduce one of them. Under usual circumstances, you and they have exactly the same replication ability: the ability to search the literature for a replication.
In my opinion, you have the Science/Nature thing vs their spinoff exactly the other way around. Science/Nature only publish flashy things more for their "memeness" value than their actual scientific content. The first driver leading you to those journal is how much the paper can be talked about and picked up by other media outlets. In fact, the most important thing you need to publish in Science/Nature is NOT a good experiment/result, but being a coauthor of somebody who already has a lot of Science/Nature papers. Their spinoffs, on the other hand, being less driven by popularity, often offer more solid content.
R00 that I typically use is about 2.1-2.6, however relating this to the 30-60% case increase per day is a difficult task.
If you know more information about where I could find these I would appreciate it.