Coronavirus Sars-CoV-2/Covid-19 Megathread

[alfred russel]

AF stop with the nonsense of using Canada as a whole as a paragon comparator.  Quebec has done very poorly.  Compare Georgia with not Quebec and see how it looks.

Quebec is like a quarter or something of your population. Can I remove the worst performing quarter of Georgia's counties as well? Or is the fair comparison between all of Georgia and the parts of Canada that are doing well?

Comparing a small geographic area to a country as vast as Canada without accounting for regional differences has some obvious deficiencies but if that is what you need to do to keep your narrative, carry on

Quebec is a larger geographic territory than Georgia and about the same in population. Is that unfair?

Canada ex. Quebec: good
Georgia: average
Quebec: bad

Canada overall: average

I don't see why you would interpret those data points as Canada being better.

[Sheilbh]

Isn't the issue that some states implemented them early and then revoked them early too?

So what's happening in the US seems to be what I thought would happen - there'd be multiple waves as it hits different bits of the country (starting with the most linked to Europe/Asia). But everyone locked down in March, regardless of whether there was much community transmission going on. So they "beat" it. Lifted lockdown and are now getting their first wave but have already suffered from lockdown which will make it more difficult to do again.

[DGuller]

Quebec is a larger geographic territory than Georgia and about the same in population. Is that unfair?

Canada ex. Quebec: good
Georgia: average
Quebec: bad

Canada overall: average

I don't see why you would interpret those data points as Canada being better.

CC has a point.  In statistics, you have to take care in what segments of data you include in the averages.  There is no general rule as to what should be included in the average and what should be excluded, but it usually becomes clear after you know what conclusion you want to arrive at.  Don't worry about justification, you'll find one that fits for whatever you ultimately decided to do.

[alfred russel]

Here is why I'm focusing on rate of growth of cases:

1)  In the long run it doesn't matter what the testing policy is.  Regardless of what share of cases you catch with diagnoses, after a one-time adjustment it will not have an effect on growth rates.
2)  Deaths follow cases, with some lag.  If your cases are growing, your deaths will be growing too in due time.
3)  Cases are more statistically credible.  Death numbers can be a little hard to work with, and in some states apparently some people come back to life due to Covid.  Metrics other than cases or death are not collected well enough on a uniform basis, so they're not even in consideration for me.
4)  There is no good scenario where case growth accelerates without the situation going unambiguously bad by all metrics, whether death or necessity of lockdown.

With regard to point 1, it will matter if testing policy is adjusted over time. For example, if you are steadily increasing the number of tests, or people administering the tests are becoming better at the task.

With regard to point 2, the relationship won't be linear. It depends on who is getting infected.

But one issue is that you are using flow rather than stock measures. Both obviously have uses, but if we are judging government response...NYC apparently has ~20% of the population infected at some point. That is going to provide a material slow down on transmission rates. It depends on how things play out, but it may be a positive to have younger healthy people infected. It is plausible that case growth accelerating among low risk segments of the population is a positive, provided that case growth doesn't transfer to high risk segments.

[Eddie Teach]

I think CC would really prefer to just compare everyone to BC.

[crazy canuck]

AF stop with the nonsense of using Canada as a whole as a paragon comparator.  Quebec has done very poorly.  Compare Georgia with not Quebec and see how it looks.

Quebec is like a quarter or something of your population. Can I remove the worst performing quarter of Georgia's counties as well? Or is the fair comparison between all of Georgia and the parts of Canada that are doing well?

Comparing a small geographic area to a country as vast as Canada without accounting for regional differences has some obvious deficiencies but if that is what you need to do to keep your narrative, carry on

Quebec is a larger geographic territory than Georgia and about the same in population. Is that unfair?

Canada ex. Quebec: good
Georgia: average
Quebec: bad

Canada overall: average

I don't see why you would interpret those data points as Canada being better.

Ok, one last try at this.  Quebec does not equal Canada.  In fact Quebec is an extreme outlier within Canada.  If you need to compare a state to Quebec to demonstrate it is doing ok, you have big problems.

[alfred russel]

CC has a point.  In statistics, you have to take care in what segments of data you include in the averages.  There is no general rule as to what should be included in the average and what should be excluded, but it usually becomes clear after you know what conclusion you want to arrive at.  Don't worry about justification, you'll find one that fits for whatever you ultimately decided to do.

I compared Canada to Georgia because CC has been touting the Canada response and comparing the country positively with specific US states.

If he wants to exclude Quebec from Canada, the burden should be on him to explain why Quebec is an outlier that should be excluded.

[Zanza]

I don't think I ever commented on Georgia, but I am one of those that were sceptical about lifting the lockdown in Germany when it was lifted. For similar reasons as those stated here for Georgia. I was so far wrong about it as luckily there is no second wave in the general population. But I think the lockdown and the following restrictions and hygiene rules were very effective. For Germany,  I have seen scientific statements to that end too.

I generally have the impression that while case numbers grow, death numbers seem to decline. Maybe they made significant progress in the clinical treatment?  Will be interesting to learn the cause eventually when scientists figured it out.

[DGuller]

I think CC would really prefer to just compare everyone to BC.

And I'm sure he'll go down to a single village in BC if he has to, but there are reasonable boundaries.  A good rule of thumb is to cherrypick just enough to get to the conclusion you need, and maybe add a little margin.  You don't want to tax your confirmation bias skills anymore than you need to, you might need the energy reserves later.

[crazy canuck]

I think CC would really prefer to just compare everyone to BC.

That would be ideal, New Zealand also works along with a few other places that did this right.  It is a bit horrifying to watch people like AF make comparisons to others that have done poorly to say they are not so bad.

[crazy canuck]

I think CC would really prefer to just compare everyone to BC.

And I'm sure he'll go down to a single village in BC if he has to, but there are reasonable boundaries.  A good rule of thumb is to cherrypick just enough to get to the conclusion you need, and maybe add a little margin.  You don't want to tax your confirmation bias skills anymore than you need to, you might need the energy reserves later.

You are a very good enabler for AF.

[DGuller]

Here is why I'm focusing on rate of growth of cases:

1)  In the long run it doesn't matter what the testing policy is.  Regardless of what share of cases you catch with diagnoses, after a one-time adjustment it will not have an effect on growth rates.
2)  Deaths follow cases, with some lag.  If your cases are growing, your deaths will be growing too in due time.
3)  Cases are more statistically credible.  Death numbers can be a little hard to work with, and in some states apparently some people come back to life due to Covid.  Metrics other than cases or death are not collected well enough on a uniform basis, so they're not even in consideration for me.
4)  There is no good scenario where case growth accelerates without the situation going unambiguously bad by all metrics, whether death or necessity of lockdown.

With regard to point 1, it will matter if testing policy is adjusted over time. For example, if you are steadily increasing the number of tests, or people administering the tests are becoming better at the task.

With regard to point 2, the relationship won't be linear. It depends on who is getting infected.

But one issue is that you are using flow rather than stock measures. Both obviously have uses, but if we are judging government response...NYC apparently has ~20% of the population infected at some point. That is going to provide a material slow down on transmission rates. It depends on how things play out, but it may be a positive to have younger healthy people infected. It is plausible that case growth accelerating among low risk segments of the population is a positive, provided that case growth doesn't transfer to high risk segments.

On point 1, it's still just a one-shot upwards adjustments that's going to peter out.  Increased testing won't sustain the growth acceleration indefinitely.  That said, I will admit to not having a quantifiable idea of how significant that distortion will be and for how long.  As for as the other points, I think they're reasonable.  I don't subscribe to them, but I can't really produce arguments to debunk them.

[crazy canuck]

Just make sure you keep using the correct account when you post Dorsey.

[DGuller]

You are a very good enabler for AF.

I know you mean it as an insult, but I take it as a compliment.  One of the saddest casualties of the Covid-19 has been the death of critical thinking.  AR somehow avoided the first wave, maybe the mountains protected him.

I disagree with AR much more than I agree with him here.  But you know what?  His arguments are much harder to debunk than other people's arguments than line up with my view of things.  One of the things that I absolutely detest is coming up with a result first and then back-fitting the analysis to get there, to me that is a mortal sin, and it's never going to lead you to a good place.  This thread is polluted with such sinners here, and many of them also aggravate their intellectual dishonesty with completely inexcusable level of bullying (I suspect partly because they always come off second-best in direct debate of facts).

[DGuller]

Just make sure you keep using the correct account when you post Dorsey.

If you're going to borrow arguments from your intellectual betters, this is a good start, but you ultimately may want to aim a little higher than Tamas.