2020 Dumpster Fire (Enter at your own risk)

[Oxbow6]

StarTribune reporting 1 in 5 in MN has lost work or is unemployed in the past 6 weeks.

[Oxbow6]

As of Wednesday 600 pm ET in NYC 34% of recorded deaths were probable (3914). Confirmed were 7563.

 

Funny thing is I still can't get the sportsbook at Caesars in Vegas to pay me on my bet on the Bills in SB XXV even though almost everyone thought Norwood was "probably" going to make that field goal attempt.

[MafiaMan]

14 minutes ago, Oxbow6 said:

StarTribune reporting 1 in 5 in MN has lost work or is unemployed in the past 6 weeks.

Well, it’s for the best, since they’re “safer at home” anyways.  Remember, “we’re all in this together.”  Well, except that government officials still have their jobs and are collecting their salaries, but other than that, you know, it’s pretty much the same.

[Nodak78]

IHME projection changes for Midwest states.

https://www.powerlineblog.com/archives/2020/04/coronavirus-in-five-states-2.php

 

[wasmania]

33 minutes ago, Nodak78 said:

IHME projection changes for Midwest states.

https://www.powerlineblog.com/archives/2020/04/coronavirus-in-five-states-2.php

 

I dont understand (rationally) why so many here are desperate to conclude that epidemiological models are a tool in a nefarious plot to keep them unemployed, cause a political downfall of their favorite candidate etc.   Bad forecasts are a hallmark of modeling things that are poorly understood or incredibly dynamic.  As the data improves the accuracy gets better.  I urge you to give equal time to sober explanations of the forecasting process (Johns Hopkins, CDC etc.) , in addition I guess to the blogs and pundits that tell you what you WANT to hear.

[jdub27]

1 hour ago, Oxbow6 said:

As of Wednesday 600 pm ET in NYC 34% of recorded deaths were probable (3914). Confirmed were 7563.

 

Funny thing is I still can't get the sportsbook at Caesars in Vegas to pay me on my bet on the Bills in SB XXV even though almost everyone thought Norwood was "probably" going to make that field goal attempt.

Your analogy attempt was more off the mark than Norwood's kick.

[yzerman19]

Remember when you’re a hammer, everything looks like a nail.  These epidemiologists have literally been studying this their whole life and are “geeked” to play with their toy models.  

[dynato]

9 minutes ago, wasmania said:

I dont understand (rationally) why so many here are desperate to conclude that epidemiological models are a tool in a nefarious plot to keep them unemployed, cause a political downfall of their favorite candidate etc.   Bad forecasts are a hallmark of modeling things that are poorly understood or incredibly dynamic.  As the data improves the accuracy gets better.  I urge you to give equal time to sober explanations of the forecasting process (Johns Hopkins, CDC etc.) , in addition I guess to the blogs and pundits that tell you what you WANT to hear.

It is easier for most to reject stats and models outright than to spend 5 minutes to learn about where they are valid and where they are invalid. Models do not hide this information, they tell you these points of contention explicitly. If they don't, then it isn't a model, it is just conjecturing and guessing. 

[Oxbow6]

7 minutes ago, jdub27 said:

Your analogy attempt was more off the mark than Norwood's kick.

Probably

[Siouxphan27]

[UNDBIZ]

1 minute ago, dynato said:

It is easier for most to reject stats and models outright than to spend 5 minutes to learn about where they are valid and where they are invalid. Models do not hide this information, they tell you these points of contention explicitly. If they don't, then it isn't a model, it is just conjecturing and guessing. 

In the age of twitter, all nuance and context are lost.  Unfortunately it goes both ways and important decisions are being made based on headlines, rather than underlying information, both in this fight and in life before the Rona.

[DeadPoets]

1 hour ago, Nodak78 said:

IHME projection changes for Midwest states.

https://www.powerlineblog.com/archives/2020/04/coronavirus-in-five-states-2.php

 

I do not know what happened to the other states, but I have a fairly good idea of what happened in ND: data entry errors.  I was checking the IMHE site regularly, and on/around 4/6 their model increased ND deaths from like 67 to around 690.  The problem appeared to be linked to actual daily death counts for 4/4 and 4/5 showing 19 and 17.  Obviously those were very wrong: ND didn't have any deaths on those dates.  I believe (although cannot say for certain) that spike caused the models to spit out the 690 number.  

I e-mailed IMHE about this and received an auto-response with links to some of the underlying assumptions and model changes.  One of the recent model changes was that the projections shown were actually the average of the 3 most recent projections.  Thus, even when they corrected the number of actual deaths for those days, it took a few projection updates to flush them out of the system fully.  I believe it went from 690 to like 575 to 389 and finally to where it is now, 32.

Obviously assumptions and methodology are the very important in the accuracy of the models, but when the data is incorrect or bad, there is no hope of the projections being accurate.  That appears to be what occurred in this instance.

[Nodak78]

25 minutes ago, DeadPoets said:

I do not know what happened to the other states, but I have a fairly good idea of what happened in ND: data entry errors.  I was checking the IMHE site regularly, and on/around 4/6 their model increased ND deaths from like 67 to around 690.  The problem appeared to be linked to actual daily death counts for 4/4 and 4/5 showing 19 and 17.  Obviously those were very wrong: ND didn't have any deaths on those dates.  I believe (although cannot say for certain) that spike caused the models to spit out the 690 number.  

I e-mailed IMHE about this and received an auto-response with links to some of the underlying assumptions and model changes.  One of the recent model changes was that the projections shown were actually the average of the 3 most recent projections.  Thus, even when they corrected the number of actual deaths for those days, it took a few projection updates to flush them out of the system fully.  I believe it went from 690 to like 575 to 389 and finally to where it is now, 32.

Obviously assumptions and methodology are the very important in the accuracy of the models, but when the data is incorrect or bad, there is no hope of the projections being accurate.  That appears to be what occurred in this instance.

Ok.  What about the changes for SD and Iowa.  Rather high projects.

[Nodak78]

1 hour ago, wasmania said:

I dont understand (rationally) why so many here are desperate to conclude that epidemiological models are a tool in a nefarious plot to keep them unemployed, cause a political downfall of their favorite candidate etc.   Bad forecasts are a hallmark of modeling things that are poorly understood or incredibly dynamic.  As the data improves the accuracy gets better.  I urge you to give equal time to sober explanations of the forecasting process (Johns Hopkins, CDC etc.) , in addition I guess to the blogs and pundits that tell you what you WANT to hear.

Here is a study for you.

https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v1.full.pdf+html

 

[wasmania]

17 minutes ago, Nodak78 said:

Here is a study for you.

https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v1.full.pdf+html

 

This one suggests way more people are infected than what has been identified through testing.  Looks more like a sober explanation but in a scientific lingo many people will be frustrated by.

[Nodak78]

8 minutes ago, wasmania said:

This one suggests way more people are infected than what has been identified through testing.  Looks more like a sober explanation but in a scientific lingo many people will be frustrated by.

And very low fatality rate.

[wasmania]

1 minute ago, Nodak78 said:

And very low fatality rate.

therefore what.... conspiracy?

[Cratter]

I think NYC needs to permanently shut down the subways so a lot of lives can be saved every year.

[MafiaMan]

4 minutes ago, Cratter said:

I think NYC needs to permanently shut down the subways so a lot of lives can be saved every year.

Well, if they can ban plastic bags and 52 ounce cups of pop...

[Nodak78]

36 minutes ago, wasmania said:

therefore what.... conspiracy?

Therefore your models are #@@)%&= 

Some states can open up with low fatalities.   protect the elderly, Nursing homes, etc.  The rest can get back to work with social distancing.   NY and NJ and a few hot spots are a different animal.

[wasmania]

4 minutes ago, Nodak78 said:

Therefore your models are #@@)%&= 

Some states can open up with low fatalities.   protect the elderly, Nursing homes, etc.  The rest can get back to work with social distancing.   NY and NJ and a few hot spots are a different animal.

See my previous posts agreeing and pointing out that models driven by unknown dynamic phenemona are usually very inaccurate until the dynamism and other unknowns are captured in the data.   So this is not a point of contention between us.  My question is why cannot this be accepted as truth rather than a rush to say someone is cooking the books for malignant purposes?   A version of Hanlon's law below:

"Never attribute to malice that which is adequately explained by stupidity"  (Stupidity meaning the ignorance inherent in early models that lack sufficient data history)

[Nodak78]

12 minutes ago, wasmania said:

See my previous posts agreeing and pointing out that models driven by unknown dynamic phenemona are usually very inaccurate until the dynamism and other unknowns are captured in the data.   So this is not a point of contention between us.  My question is why cannot this be accepted as truth rather than a rush to say someone is cooking the books for malignant purposes?   A version of Hanlon's law below:

"Never attribute to malice that which is adequately explained by stupidity"  (Stupidity meaning the ignorance inherent in early models that lack sufficient data history)

My point is the models didn't reflecting NYC subway transportation and people stacked in apartment to the sky vs living on the prairie and driving a pick and waving at someone.  Some elites don't know the different and how it affects community spread.

[Walsh Hall]

10 minutes ago, wasmania said:

See my previous posts agreeing and pointing out that models driven by unknown dynamic phenemona are usually very inaccurate until the dynamism and other unknowns are captured in the data.   So this is not a point of contention between us.  My question is why cannot this be accepted as truth rather than a rush to say someone is cooking the books for malignant purposes?   A version of Hanlon's law below:

"Never attribute to malice that which is adequately explained by stupidity"  (Stupidity meaning the ignorance inherent in early models that lack sufficient data history)

There can be a very fine line between malice, ignorance, and stupidity.  It seems, and seemed, completely incomprehensible that the denominator is/was only the folks who have tested positive.  Basing the fatality rate and models on the confirmed cases, or low multiples thereof, could very well be slanted more toward malice than ignorance.  I find it hard to believe that the architects of the models are stupid, and they should be well aware of what they don't know.

Let's hope the denominator is consist with the study posted above.

[wasmania]

Just now, Walsh Hall said:

There can be a very fine line between malice, ignorance, and stupidity.  It seems, and seemed, completely incomprehensible that the denominator is/was only the folks who have tested positive.  Basing the fatality rate and models on the confirmed cases, or low multiples thereof, could very well be slanted more toward malice than ignorance.  I find it hard to believe that the architects of the models are stupid, and they should be well aware of what they don't know.

Let's hope the denominator is consist with the study posted above.

stupid is the word used in Hanlon's law, I suggested ignorant as more accurate in this case.  When the only data you have comes from early spread in high density areas, it wont do well predicting low density outcomes if the model is data starved.  It was learned very late that asymptomatic people were big carriers.  If this was known earlier, than  a variable like % of people who daily use public transportation would probably be huge in the model and result in low forecasts for places like North Dakota.  Not malicious, ignorant of a key dimension.

[Nodak78]

https://www.powerlineblog.com/archives/2020/04/the-pandemic-where-we-stand-now.php