Did you catch it ?

chemgeek said:
Both the US overall or NY state data has not grown linearly. You can see that by simply examining a full plot of the raw data. It's not a straight line. Both caseload and deaths data has closely followed a canonical logistic curve, which is a self-limiting exponential growth function typical of phenomena like epidemics.
Click to expand...

No of course the confirmed cases are not a straight line overall. The question is, if you look at this on a log scale, once there were a reasonable number of cases it appeared to be on an exponential growth curve. Then for the last 4 weeks it has looked quite linear. See attached.
TotalCovidCases1.png


When I attempt to fit the total number of cases with a 5 parameter log-logistic model in R the fit for the same datapoints looks like the following (sorry about the sloppy scale labelling):

LogisticFit1.png


So I am wondering how one distinguishes a case where there is some underlying linear growth factor which has taken over from the case of a log-logistic model with a higher maximum so the curve doesn't turn over like that yet? I guess there is likely a way to make such a model comparison in R, but practically, at this point would one be able to distinguish those two alternatives in a statistically significant manner? It seems like the data may just not really be able to distinguish this but the normal course of epidemics would suggest there is a higher limit involved here where it will turn over. I guess I have to go read the IHME model...
 
I have now read the IHME model paper as well as the description of the recent updates. A few observations.

This is an empirical model which basically fits curves to the observed growth rates relative to different interventions and then uses those curve fits to predict the results in other places and for future times.

The model uses a cumulative Gaussian error function to estimate case numbers as a function of time, not a log-logistic. (EDIT: Thinking about this more I am now a bit concerned about this. The cumulative Gaussian error function is symmetric about the middle and the roll up toward the middle and the roll off at the top are mirror images. The log-logistic function is rather different in that the roll off at the top tends to be more extended. This might result in making poor predictions about the tail end behavior - which is exactly where people have complained about the predictions of this model seeming unrealistic.)

The Wuhan data figured heavily in their initial formulation, though they have many other data points as well. Of course the Wuhan data is highly suspect but it is unclear how much of an effect overall this would have on their results.

They found that the overload of hospital capacity in the initial report was 25% for ICU beds and 7% of hospital beds.

It strikes me as a fairly good job of curve fitting. I have not seen yet a comparison of model outcomes to true predictions. We know recent revisions resulted in a 25% decrease in predicted deaths. Some of their predictions in the first iteration were clearly off, for example, that New York deaths would peak in the first week of April. I don’t think we really know yet how predictive this model is, especially not in quantitative terms.

I don’t think this modeling approach says much regarding whether social distancing measures, voluntary or coercive, have had an effect on the growth rate of cases or the magnitude of such effects. That was not really its goal and the authors assume that. Likely the data and some parts of the model could be used to explore that question.
 
Last edited:
IK04 said:
I wasn't talking about medical "experts," but they are also most certainly swayed by political influences.
Click to expand...
Just because you can imagine something doesn't make it true (let alone "most certainly" true, LOL).
 
How do we avoid a long period of bouncing between doing what's best for the economy and doing what's best for getting the pandemic under control? This article suggests that we need to pay more attention to how it has been handled in countries that have gotten the epidemic under control relatively quickly, and discusses what it's going to take for the U.S. to deal with this successfully.

It’s Not Too Late to Go on Offense Against the Coronavirus

"People who have fought epidemics in the past know the reality: there is no way out of our predicament but to build a full, five-part response and keep it in place for the foreseeable future."​
 
Palmpilot said:
How do we avoid a long period of bouncing between doing what's best for the economy and doing what's best for getting the pandemic under control? This article suggests that we need to pay more attention to how it has been handled in countries that have gotten the epidemic under control relatively quickly, and discusses what it's going to take for the U.S. to deal with this successfully.

It’s Not Too Late to Go on Offense Against the Coronavirus

"People who have fought epidemics in the past know the reality: there is no way out of our predicament but to build a full, five-part response and keep it in place for the foreseeable future."​
Click to expand...

That's a good read
 
Gary said:
When the the American public has the confidence to venture out and freely associate with others. Who knows when that may be. Any further explanation, would probably launch this into a locked thread.
Click to expand...
I don't think it will be so much of an "all-clear" but rather a "little bit more clear yesterday, little bit more clear today, little bit more clear tomorrow" type of thing.
 
Palmpilot said:
The idea that the medical folks are basing their statements on politics is pure fantasy.
Click to expand...
Perhaps, but the idea that the news coverage of what the medical folks are saying isn't based on politics is certainly fantasy.
 
Dana said:
Perhaps, but the idea that the news coverage of what the medical folks are saying isn't based on politics is certainly fantasy.
Click to expand...
Not much opportunity to distort what they're saying when it comes to live interviews of the medical folks. And they have the same access to the Internet as anyone else, in case they feel they're being misrepresented.
 
PeterNSteinmetz said:
Curious, what did you think of the controls which were run by the manufacturer of the kits used in the Santa Clara study? They tested against 371 pre-Covid samples with 369 negative results. It strikes me that argues that that test is not reacting with other coronaviruses which were around at the time, though I suppose depending on where the pre-Covid samples were obtained, they might not have contained the other coronaviruses.

Article available here: https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v1
Click to expand...

The manufacturer's notes on almost all of the the test kits actually caution that there is likely significant cross-reactiviry with other strains of Coronavirus. That gives me pause, especially given my knowledge of how ELISA tests work.
 
Dana said:
Perhaps, but the idea that the news coverage of what the medical folks are saying isn't based on politics is certainly fantasy.
Click to expand...

I think generally the scientific community is frustrated that folks are not listening. Being part of that community, I can say that overwhelmingly the scientific community is dedicated to providing the best information possible from the available data, guided by clear ethical standards. Sometimes the facts are just what we wish them to be.
 
IK04 said:
I wasn't talking about medical "experts," but they are also most certainly swayed by political influences.
Click to expand...

I will comment a bit on this question having observed and been part of the biomedical community for most of my life now.

An unfortunate trend in science today is the need for people trying to establish their careers to come up with results that can be published in high impact factor journals. This is how one obtains grant funding and obtains tenure. That need is causing people to be in more of a hurry to publish a spectacular result, which leads to people often not paying as much attention to proper controls, etc.

With respect to Covid-19, given the nature of the emergency, I think most people are pretty focused on trying to come up with solutions and provide accurate information about this pandemic. However, given the nature of the emergency, a lot of things are having to move very quickly.

I do not think this means that scientists will bias the results to support one political leaning or the other. What these pressures do is bias scientists to try and report a spectacular result as quickly as possible. And of course, even without these pressures, one naturally wants one’s work and results to be exciting and important.

Since the Federal government now funds a vast majority of biomedical research, I think there is an implicit bias which is acquired by most researchers over the course of their careers.

They naturally think that big government spending, particularly on the NIH and biomedical research is a good thing. I mean the average working principal investigator probably spends 60% of more of their time on seeking grant funding (it is a constant complaint that there is no time to actually do science).

This also tends to produce professional scientists who think that government action is the solution to many problems - that is just naturally the way they think things work given the environment they are working in.

With respect to policy, most epidemiologists and medical professionals are narrowly focused on preventing a disease, and not necessarily well trained to consider collateral damage from their policies in terms of deaths due to other causes or economic impacts. I think economists have actually been doing a better job of considering those angles.

Overall, I think a big problem in the present circumstances is there is just a lot we frankly don’t know. And that is not a good or comfortable position from which to make policy.
 
Kenny Phillips said:
So "medical folks" are immune to politics? Pure fantasy.
Click to expand...
Not immune, and there are always exceptions, but I've seen a lot of of interviews with them and other public statements, and I'm not seeing the evidence that it's a widespread problem.
 
Last edited:
PeterNSteinmetz said:
So I am wondering how one distinguishes a case where there is some underlying linear growth factor which has taken over from the case of a log-logistic model with a higher maximum so the curve doesn't turn over like that yet? I guess there is likely a way to make such a model comparison in R, but practically, at this point would one be able to distinguish those two alternatives in a statistically significant manner? It seems like the data may just not really be able to distinguish this but the normal course of epidemics would suggest there is a higher limit involved here where it will turn over. I guess I have to go read the IHME model...
Click to expand...

You will see linear growth when the effective R0 value is near 1.0. In a canonical uncontrolled epidemic, the effective R0 changes because the susceptible population decreases as more and more individuals are infected. At the midway point of the epidemic, when 50% of the total that will be infected is reached, the effective R0 value gets to one and growth briefly looks linear, before an exponential decline in infections occurs.

In our current pandemic, R0 is not declining because of herd immunity--we will know more exact numbers soon for selected regions of the nation, but the statistics suggest that the penetration even in hard-hit areas is not much more than 10% of the population--but because we are controlling transmission by reducing interpersonal contact. In many regions of the country, the effective R0 is close to 1.0, and indeed for the country as a whole that is pretty much where it is, with about as many states with effective R0>1 as states with R0<1. As long as effective R0 remains close to 1.0, growth will continue unabated at a constant rate. It's really that simple.

The key to staunching the epidemic is to control the effective R0. There are limited ways of doing that, and until a vaccine is deployed, they are all unpleasant choices in one way or another, but they all require a method of interrupting transmission, and hence decreasing effective R0. Left to its own devices, the unconstrained R0 for SARS-CoV-2 is in the 2-3 range, and we know what that does.
 
Here is a serology primer for those of you interested in antibody testing. I think it is important to understand how the most popular methods work so that there is better understanding of what they can tell you. Maybe you will find this TLDR. I hope not, so you will be more informed. Here goes...

The most prominent technology being used to test for COVID-19 antibodies is an Enzyme Linked ImmunoSorbent Assay (ELISA). It is a clever technique, invented many decades ago, that depends on (hopefully) specific antibody-antigen interactions. There are two basic methods, illustrated here.

Direct assay. Many of IgM assays use this method. (IgM is the initial batch of antibodies you make after infection.) A plastic well plate is coated with anti-IgM antibodies. A serum sample from a patient is added to the wells, and any IgM in it is bound by the anti-IgM antibodies. Then a COVID-19 protein attached to an enzyme (usually horseradish peroxidase) is added to the well plate. If there are IgM antibodies bound to the plate that have affinity for the COVID-19 antigen target, they stick together. Now you add a chemical that horseradish peroxidase will react with to make it change color. You measure the intensity of the color to determine if, or in some cases, how much IgM antibody was present.

Indirect assay. The IgG assays typically use this method (IgG is the longer term antibodies that you make about 2 weeks after infection.) A plastic well plate is coated with a COVID-19 target protein (antigen). A serum sample from the patient is added to the wells, and any antibodies (not just IgG, but any antibody) that have affinity for the target protein stick. Now you add an anti-IgG antibody that is attached to an enzyme (usually horseradish peroxidase). Any IgG that happened to bind to the target antigen will now bind to the angi-IgG-horseradish peroxidase, creating a three layer sandwich. Now you add the color reagent which the horseradish peroxidase reacts with to produce a color that can be measured.

So what does this method tell you? It tells you if there are any antibodies in an individual's serum that has affinity for the COVID-19 target (antigen) that was selected for the assay. No more, no less. The target protein (or portion thereof) must be carefully chosen to prevent false positives and minimize non-specific antibody-antigen interactions. For COVID-19, the most popular targets are the spike protein (S) and the nucleocapsid protein (N). There is more N protein in the virus than S, so some manufacturers choose it for more sensitivity, but the S protein may be more unique to COVID-19. However, both S and N share significant similarity in protein structure (sequence identity) to other common strains of coronaviruses (that cause the common cold), so there is a very real, and as yet not well understood, problem in specificity. There could be a very significant fraction of false positives if an individual has recently recovered from a common cold. In fact, almost every kit manufacturer cautions users about this issue. The test kit manufacturers HAVE tested their kits against other virus antibodies or antigens (depending on the test type) that are significantly different from COVID-19 but may cause similar symptoms, e.g. influenza (various types), RSV, hepatitis. The validation studies for these targets show good discrimination.

Another complication is that not all anti-COVID-19 antibodies will bind well to the selected antigen in the test kit. Your immune system does not choose any one particular COVID-19 target, much less a specific piece of a target of COVID-19. This problem is the reason for false negatives, which in some cases can be very significant.

So what should you take away from the serology tests? If you are sero-positive, then you very likely have antibodies that can bind to some portion of the COVID-19 virus. It does not necessarily mean that you were infected previously with SARS-CoV-2, although if you were you will likely test positive, subject to the caveat about false negatives above. The good news is that it MAY mean you have some degree of immunity to SARS-CoV-2, even if you were not actually infected with it. (Count your lucky stars that one of your previous colds provided antibodies that were "close enough" to confer some resistance.) If you were infected with SARS-COV-2, the ELISA test will PROBABLY pick up that fact. So, bottom line, a positive test is PROBABLY good news for you. But note that the test is not necessarily a great tool for measuring how many individuals were actually infected in the past, depending on how bad the false positive rate is. And that false positive rate will be different for every manufacturer that uses a different target COVID-19 antigen protein. In the rush to get these tests out, it is not clear at this point that the cross-reactivity issues with other coronaviruses has been thoroughly explored. The kits are primarily being used to confirm diagnoses of COVID-19, so the cross-reactivity and false positives are not as large an issue for that purpose, because in that testing population there will be few true negatives. However, for surveying a broader population, the false positive rate, even a modest one, is big deal, because maybe 90-98% of the measurements are expected to be negative. So a small false positive rate in a large population of negative subjects will significantly distort the apparent data.
 
As some of you know, the recent random serology testing done in NY state suggests a 14% overall exposure rate for the general population. This does inform us of a few things, with a few caveats. First, the overall high rate means that the false positive distortion of the data is fairly low. For reference, the serology tests, AT BEST, have somewhere between a 0.5 and 2.0% false positive rate. I don't think any of the manufacturers have thoroughly investigated the cross-reactivity with other, harmless, commonly found coronoviruses. But I digress. Assuming the reported false positive rates of 0.5-2.0% are accurate, one can mathematically show that these are not significant once you get beyond a 10% positive rate in your sample. However, one should be cautioned that the exposure rates varied from about 20% in NY City to around 3-4% in Western NY. At 3-4% positive rates in the sample population, false positives will inflate the true number by 8-50% (relative), depending on the exact false positive rate and true positive rate. So the overall exposure rate in NY state is probably slightly inflated, but likely not egregiously so. And this study, unlike the one in Santa Clara County, CA was closer to a random sample, so there is likely less sampling bias toward those with a high likelihood of exposure.

So, all told, we have a rough idea of the infection fatality rate (IFR) in NY state, and if you do the math, it's around 0.57%. Additional caveats: this is probably a slight underestimate, because it does not account for COVID-19 deaths that did not occur in a hospital or other setting where the deceased would be identified as COVID-19 deaths.It also may or may not be a sample that is well-distributed in terms of age or gender. In addition (see above), the exposure rate is probably slightly overestimated. But moving forward anyway with the 0.57% number, the IFR is still at least 3-6X that of seasonal flu. So, not the 5% that we see in the case fatality rate (CFR) in the US, but still pretty serious. BTW, a 5% CFR is near the international median now. If you get sick enough with this virus to seek treatment, you have a 1 in 20 chance of a really bad result. Fortunately, most who get COVID-19 apparently don't know it, or just assumed they had a bad cold or common flu. (I've certainly had enough colds this winter season--like every time I come back from commercial airline travel.)

The measured exposure rates were slightly higher than that expected by most scientists and medical professionals (I think the over-under was probably around 10%), but not that far off, either. Herd immunity is something around 70% or higher, depending on the exact R0 for SARS-CoV-2. We won't get there until several waves of infection like the current one (too painful to consider), or we intervene with a vaccine. The next stage will be to keep outbreaks to a containable slow boil until such time as we can depoly a vaccine. If we can deploy some effective therapeutics in the mean time (I'm looking at YOU, RNA-dependent RNA polymerase inhibitors!), we have a good chance at moving forward to a more normal existence.
 
No empirical evidence for the efficacy of coercive lockdowns. This is exactly what I have been saying for some time. EDIT: I obtained the dataset and verified his results. Of the independent variables he explored, population and its density have a statistically significant effect on the cases and deaths, the adoption of a coercive lockdown has none, not even close.

https://www.spiked-online.com/2020/04/22/there-is-no-empirical-evidence-for-these-lockdowns/
 
Last edited:
chemgeek said:
So, all told, we have a rough idea of the infection fatality rate (IFR) in NY state, and if you do the math, it's around 0.57%.
Click to expand...

Also note that estimates in other locations are lower, such as 0.3% in the Gangelt Germany study and 0.12-0.2% in the Santa Clara county study. There may well be other variables which determine whether people die from Covid-19 after being infected. If we can figure out what those are, another opportunity to decrease the fatality rate.
 
Last edited:
Palmpilot said:
You just LOVE that word "coercive"!
Click to expand...

It is an important distinction I think in all areas of political policy. If people want to do something, like stay at home, voluntarily, that is up to them. If you have to coerce them into doing something, that is a very different question.

What is important in this study is that there is no empirical evidence the lockdown orders for Covid-19 have slowed the spread of disease. That is really extremely important given the collateral damage in terms of other deaths and economic losses which are caused, at least in part, by these coercive orders.

I do think there would be some economic damage without coercive orders in any case as people socially distance voluntarily, however, it would likely be more tailored and limited.
 
PeterNSteinmetz said:
... there is no empirical evidence the lockdown orders for Covid-19 have slowed the spread of disease...
Click to expand...
If it didn't, then the disease spreads magically, or perhaps we aren't 'locked down' tightly enough. A true lockdown, with very limited group size, would have already killed this thing.
But think if it the other way: if you give the disease to someone, should you be able to be sued?
 
Kenny Phillips said:
If it didn't, then the disease spreads magically, or perhaps we aren't 'locked down' tightly enough. A true lockdown, with very limited group size, would have already killed this thing.
But think if it the other way: if you give the disease to someone, should you be able to be sued?
Click to expand...

No question that extreme social isolation would work in the limit. But in the real world of the US, there may be a lot of reasons, nothing magical, why the coercive lockdowns would not have an effect.

Could be there is little compliance in areas where it matters. Or that voluntary compliance is sufficient. To study that one likely needs to correlate actual measures of social mobility and interaction with the spread of disease.

The question about lawsuits is an interesting one. Is it a tort if you know you have infectious Covid-19 and travel in public? Seems to me that will depend on the likelihood that by doing so you harm someone else. And that is some sort of product of the likelihood you infect someone else and then the chance of them being seriously injured or hurt by being infected. We know a little bit about the latter now and less about the former.
 
Last edited:
PeterNSteinmetz said:
It is an important distinction I think in all areas of political policy. If people want to do something, like stay at home, voluntarily, that is up to them. If you have to coerce them into doing something, that is a very different question.

What is important in this study is that there is no empirical evidence the lockdown orders for Covid-19 have slowed the spread of disease. That is really extremely important given the collateral damage in terms of other deaths and economic losses which are caused, at least in part, by these coercive orders.

I do think there would be some economic damage without coercive orders in any case as people socially distance voluntarily, however, it would likely be more tailored and limited.
Click to expand...

It didn't look to me like that "study" was published in a scientific journal, and its author is political scientist, not a member of the life sciences.
 
Palmpilot said:
It didn't look to me like that "study" was published in a scientific journal, and its author is political scientist, not a member of the life sciences.
Click to expand...

No, it is not published in a peer-reviewed journal so like a pre-print the normal cautions apply. One can argue endlessly about the credentials of a political scientist, social scientist, or epidemiologist to evaluate something like this and their potential biases. Sort of an ad hominem attack on the argument he is making.

What did you think of his argument and statistical analysis? I obtained his dataset and the basic statistical results from that data check out.

And please note, at this stage I am not aware of any empirical studies performed on Covid-19 in the US which suggest that coercive measures have had an impact on the spread of the illness -- are you aware of any? If so, please post the links.
 
Last edited:
PeterNSteinmetz said:
What did you think of his argument and statistical analysis? I obtained his dataset and the basic statistical results from that data check out.
Click to expand...
His statistical analysis may or may not be valid, but I saw several errors and omissions in the data he based it on. His states without lockdown orders is far from complete, he doesn't allow for the wide variance even within a single state (there is a huge difference, for example, between New York City and the rural counties only a hundred miles north of the city), and the variable of time (NY got hit early while the more rural areas haven't been hit yet to any great extent).

His conclusions may or may not be correct, I don't have the data either, even a random guess may or may not be correct.
 
PeterNSteinmetz said:
No, it is not published in a peer-reviewed journal so like a pre-print the normal cautions apply. One can argue endlessly about the credentials of a political scientist, social scientist, or epidemiologist to evaluate something like this and their potential biases. Sort of an ad hominem attack on the argument he is making.

What did you think of his argument and statistical analysis? I obtained his dataset and the basic statistical results from that data check out.

And please note, at this stage I am not aware of any empirical studies performed on Covid-19 in the US which suggest that coercive measures have had an impact on the spread of the illness -- are you aware of any? If so, please post the links.
Click to expand...
I'm not aware of any studies that say that such mandatory measures don't work. "Absence of evidence is not evidence of absence."
 
Dana said:
His statistical analysis may or may not be valid, but I saw several errors and omissions in the data he based it on. His states without lockdown orders is far from complete, he doesn't allow for the wide variance even within a single state (there is a huge difference, for example, between New York City and the rural counties only a hundred miles north of the city), and the variable of time (NY got hit early while the more rural areas haven't been hit yet to any great extent).

His conclusions may or may not be correct, I don't have the data either, even a random guess may or may not be correct.
Click to expand...

His statistical analysis is correct - I have verified it myself using his data. (Feel free to point out any errors in it you find. He will happily send you the dataset if you ask.)

The same conclusions hold when using the latest Covid-19 tracking project data available at https://github.com/COVID19Tracking/covid-tracking-data .

In terms of the list of states without lockdown orders, that is important as it is his main independent variable. He has 8 states in the dataset without a lockdown and just using voluntary social distancing, AK, IA, NE, ND, SC, SD, UT, WY. This agrees exactly with the dataset previously posted by ChemGeek. Do you see some other problem with that?

He acknowledges that a county by county study would likely be more accurate in terms of population density. And that trying to correlate with some type of variable for the start of lockdown would be an improvement (though he also looked internationally at other countries).

So I think a fair statement is that the one study to examine this question so far failed to find evidence that coercive lockdowns have any effect on the spread of Covid-19. Future studies might. Please link to them or develop them. The rest is speculation on what might be true.
 
Palmpilot said:
I'm not aware of any studies that say that such mandatory measures don't work. "Absence of evidence is not evidence of absence."
Click to expand...

Attempting to reverse the burden of proof. (And I think you must mean "any other studies".) In general the burden of proof is on he who asserts the existence of something, not the other way around (which leads to all sorts of contradictory outcomes). Those who assert that coercive lockdowns help with suppressing the spread of Covid-19 have the burden of proving that it in fact suppresses it.

The only reason to think they might work is that in the limit extreme social distancing has to work. But extreme social distancing and coercive lockdown orders as implemented in the US are very different things and there are plenty of reasons to think coercive lockdown orders might not work. There is also no available study, analysis or data to suggest coercive lockdowns have worked in the US for Covid-19.

It is of course strictly true that failing to disprove the null hypothesis (as in this study) does not prove the null hypothesis is absolutely true.

Nonetheless, I think it is fair to say that the balance of evidence presently is that coercive lockdowns are not achieving their intended effect.

I am however continuing to develop a more sophisticated test that would include time of implementation and the full time course of the cases and deaths to see if that might show some sort of effect of the coercive lockdowns.
 
Last edited:
PeterNSteinmetz said:
Also note that estimates in other locations are lower, such as 0.3% in the Gangelt Germany study and 0.12-0.2% in the Santa Clara county study. There may well be other variables which determine whether people die from Covid-19 after being infected. If we can figure out what those are, another opportunity to decrease the fatality rate.
Click to expand...

The Santa Clara study, which has not yet been peer-reviewed IIRC, has an issue in subject selection, which was not randomized, but rather potentially biased toward individuals who may have been at higher risk of exposure. (Subjects were solicited for the study via social media.) The NY study attempted to collect a random sample (pretty much everyone it could convince at unannounced locations all around the state), but word got out at some of these locations, and people lined up to be tested, which could have biased the sample toward those with higher risk of exposure. Nevertheless, as a preliminary study, it has provided some helpful understanding. The Santa Clara study, if generalized to other communities, would predict exposure rates that are not reasonable or conforming with actual data.
 
PeterNSteinmetz said:
No empirical evidence for the efficacy of coercive lockdowns. This is exactly what I have been saying for some time. EDIT: I obtained the dataset and verified his results. Of the independent variables he explored, population and its density have a statistically significant effect on the cases and deaths, the adoption of a coercive lockdown has none, not even close.

https://www.spiked-online.com/2020/04/22/there-is-no-empirical-evidence-for-these-lockdowns/
Click to expand...

This is a pretty unconvincing argument. Doing a simple regression analysis of current case data is quite simplistic, as each state is not at the same point in their epidemic curve. NY state is past peak, some states are near peak, while others are not yet at peak. It is kind of like comparing baseball scores from different innings of games and then trying to make a conclusion about the differences between baseball teams.

If you really want to understand how regions are doing without testing bias, one good way is to look at the growth rates of COVID deaths. Unlike confirmed cases, COVID deaths are not as sensitive to testing rates, have a better chance of being correctly identified, and are less likely to be missed. It's fairly easy to track the apparent doubling time (or decay time) of COVID deaths with simple math. Deaths are a lagging indicator, with a delay of about 14-28 days from an policy changes that might affect infection rates. (That's how long it typically takes a individual who is exposed to SARS-CoV-2 to become a fastality.) Unfortunately, some states are doing so little testing, it is difficult to know if they are doing well or not by looking at confirmed cases. But the death growth rates can be informative.

I'm tracking confirmed cases, new cases per day, and deaths for my local community and region, and keeping up with some key states where friends and family are located. Some of those states will be interesting laboratories for early relaxation of distancing policies, even as I am concerned for my friends and family there.
 
What will be interesting is how the final numbers of the COVID plan Sweden used stack up against other countries that enforced stricter quarantine type plans. While the Swedish COVID death rate is at a higher per capita rate it still seems to follow the same mortality demographics (65+ w/ underlying health issues) as other countries. It should prove a valid benchmark when comparing the herd immunity vs isolation arguments, and by extension, the effects of both plans on a state economy. Time will tell.
 
chemgeek said:
This is a pretty unconvincing argument. Doing a simple regression analysis of current case data is quite simplistic, as each state is not at the same point in their epidemic curve. ...

If you really want to understand how regions are doing without testing bias, one good way is to look at the growth rates of COVID deaths.
Click to expand...

Reilly also studied deaths and this is one of the points he makes. Same result for total deaths. And yes, he also notes that a temporal analysis would be more informative.

It turns out that some of the non-lockdown states were infected earlier and are further along in the epidemic curves. It needs further quantitative study I think to be clearer.

But this is the best we have out in pre-print form presently.
 
Last edited:
PeterNSteinmetz said:
It turns out that some of the non-lockdown states were infected earlier and are further along in the epidemic curves. It needs further quantitative study I think to be clearer.
Click to expand...
True enough. So there are two ways to do that. You can assume the worst and clamp down with the possible worst case end result being you were completely wrong and now some people are poorer because of it. Or you assume the best and open up everything with the possible worst case end result being you were completely wrong and now some people are dead because of it.

So that's really the bottom line question in all of this. Do you want to err on the side of people being poor or do you want to err on the side of people being dead. I can definitely understand wanting to err on the side of people being dead because let's face it, dead people don't complain nearly as much as people being poor. And in the end, isn't fewer complaints all any of us are really looking for?
 
Old Thread: Hello . There have been no replies in this thread for 365 days.
Content in this thread may no longer be relevant.
Perhaps it would be better to start a new thread instead.
Back
Top