Thursday, April 09, 2020

Odds And Ends, Modeling And Branding Edition

Yesterday's post elicited typically excellent comments from Ohioan, Foxie and Tim. Here are some of the thoughts they catalyzed, certainly indicative of a deranged mind.
  • The IHME model was deliberately conservative. It had to be. If they came out predicting 20,000 deaths and we ended up with 500,000 because everyone dismissed the Chicom Flu as a non-event, they'd be culpable morally, if not in actuality.
  • If there was a cultural failing, it was to have trusted anything from China at all. First off, China is one, massive governmental state. Government employees maximize self-preservation, not profits and if that means they need to tell lies, then they tell lies. That's how the Chinese ended up with Mao's body count.

    The current Chinese government is a direct, lineal descendant of the one that demanded that peasants smelt iron in small, backyard furnaces. All of the functionaries lied about the consequences and as a result, 80,000,000 died. Like that mattered. They were all just meatbags anyway.
  • Hobby Horse Bonus Point: The individual is always a better decision maker than the government. We've lost that concept. See also: government, massive expansion of. Trusting the Chinese is a symptom of a larger failing in our culture.
  • Ohioan made my branding point quite well. As to NBC News, I have come to conclude that the acronym has changed from National Broadcasting Corporation News to Nothing But China's News. They wrecked their brand and he's closed his mind to them and moved on to other things. Recovering lost trust with your customers is a Herculean task.
  • The American model of Federalism is genius. Foxie pointed out why. *looks at model graphic* Oh. So that is why Iowa only going somewhat overboard is suddenly functionally the same as "do not leave your house unless absolutely necessary". Iowa is not New York and the people of New York shouldn't get a say in how it's run. See also: peasants smelting iron, Chinese policy of.
  • Tim weighed in with a comment that made me think of the loss of expensive, experienced journalists. I wish people reporting on these projections would spend more time talking about error bounds and intrinsic uncertainty, instead of just acting as if the mean value is the One True Value. As the revenue at the news media companies has dropped, they've fired the expensive old hands and replaced them with gullible, ideologue children. This is how we get clown show press conferences.
  • I'm going to disagree with Tim on this one, but only in a very lukewarm way. In general, they are fighting with the same issues that people who try to model the economy or the stock market run into: When they make predictions, people change their behavior based on the predictions. I think the lousy models were more due to lousy input data. All we had were Chicom lies and unverifiable information from the Italians. If you're looking to the Italians to give you an organized summary of results, you deserve what you get. ;-)
  • On the other hand, I agree with Tim's point and it reinforces my point that Federalism wins. Tell Americans what's going on and they will usually make good choices.
  • Also from Tim: Oh, and the quality of the data they are working with is evidently terrible, missing somewhere between 20% and 80% of infections, and most diagnoses being of people that were infected a week or two previously. This is at least as bad as trying to predict the stock market, and worse than predicting the weather. I'm so glad I was able to change his mind. :-)
  • Foxie then goes on to make the point that interests me the most. Of course, lying to manipulate people into doing what they would never do if given full information is why people ignore the modelers. The news media has annihilated their brand. Lost customers who have alternatives they trust aren't coming back without a practically miraculous turnaround.
One final thought. I am highly suspicious of the people who want to keep this lockdown going for more than a month longer. It seems to me that they're the same people who long to recreate Mao's Paradise. They love control and power. It sure looks to me like they're anxious to use this crisis to grab it.

Kind of like the Global Warming Climate Change thing.

9 comments:

tim eisele said...

As far as keeping the lockdown going, my big issue is that there are wildly different degrees of risk, not just between jobs but between different parts of the same job, but everything is just lumped together.

In my case, on the one hand lecturing to 50 or so engineering undergrads crammed into a room is obviously not a great idea if you want to avoid spreading diseases. But on the other hand, when I work in the lab on research projects I generally work alone or with maybe a graduate student around, sometimes don't see more than one or two other people all day, and it is a good idea for me to wear a dust mask and rubber gloves in there anyway. Lab work strikes me as almost as low-risk as staying home. Yet the university administrators make no particular distinction, and have kicked us all off of campus altogether. I am seriously considering the value of building my own lab at home, rather than letting the university control whether I can work or not.

At the other extreme, my wife is self-employed, and the authorities have no problem with her going out, because she is a "key worker" . . .

. . . she's a locksmith.

Tom said...

Regarding the desire to have a statistics discussion, look at Hurricane tracking. The NWS has been hammering on “the cone” and not including the “line”, and it seems to be making a difference (though you’ll always have those who fixate on one prediction in a spaghetti plot...).

Ohioan@Heart said...

KT. You are right, I don’t trust the modelers. That’s part of why I made my own. For it, I know precisely where the assumptions are, and how they affect the output. The ones reported by the media are black boxes with little to no insight on the parameters and assumptions, And they are filtered and garbled by people who can’t even spell model if you spot them the first four letters. Trying to extract useful information from those is impossible.

But much more importantly: the model allowed me to understand the delay between the numbers they measure and the actual cases. This tells me that we see the cases start dropping there is still a huge number of cases, and they only go away *slowly* over another month or two. This means that when the numbers start improving it is *NOT* time to loosen the restrictions, they must be maintained for two to four more weeks. Don’t do that and we all better be ready for “CoViD-19, The Sequel”. And I know you don’t think that I’m a person “who long[s] to recreate Mao's Paradise”.

Ohioan@Heart said...

KT. You are right, I don’t trust the modelers. That’s part of why I made my own. For it, I know precisely where the assumptions are, and how they affect the output. The ones reported by the media are black boxes with little to no insight on the parameters and assumptions, And they are filtered and garbled by people who can’t even spell model if you spot them the first four letters. Trying to extract useful information from those is impossible.

But much more importantly: the model allowed me to understand the delay between the numbers they measure and the actual cases. This tells me that we see the cases start dropping there is still a huge number of cases, and they only go away *slowly* over another month or two. This means that when the numbers start improving it is *NOT* time to loosen the restrictions, they must be maintained for two to four more weeks. Don’t do that and we all better be ready for “CoViD-19, The Sequel”. And I know you don’t think that I’m a person “who long[s] to recreate Mao's Paradise”.

Foxfier said...

NYC backwards -admitted that it's been reporting unconfirmed kung flu cases in their death counts-- as the result of a news story on how 911 guys are not taking cardiac arrest victims who can't be revived on site into the hospital.
They're now officially counting anybody who dies at home and has signs that could suggest kung flu symptoms as a COVID-19 death.

https://gothamist.com/news/surge-number-new-yorkers-dying-home-officials-suspect-undercount-covid-19-related-deaths

Which got me curious-- it would explain how a third of the deaths are over 80 ( https://www.syracuse.com/coronavirus-ny/ ), if all you need for a 'diagnosis' is cold, flu, or pneumonia symptoms.
Over 65 is about 17% of the population, the biggest 5 year chunk of which is 65-70;
https://www.cia.gov/library/publications/resources/the-world-factbook/attachments/images/large/US_popgraph2020.JPG

Ohioan@Heart said...

Foxy - San Diego County doesn’t have the no transport rules. They do report some interesting numbers. As of today’s numbers (considered correct through yesterday) the percent of cases for 80+ are 92 of 1693 (5.4%) while that group is 22 deaths out of 44 (50%). Even with the uncertainty in NYC, I don’t doubt for a moment that that age group could represent 1/3.

The SD data are available at
https://www.sandiegocounty.gov/content/sdc/hhsa/
programs/phs/community_epidemiology/dc/2019-nCoV/status.html
And links therein.

Foxfier said...

Ohioan-
it's less the no-transport (which is just horrifying on its own, especially since docs in area have been commenting on how their heart and stroke wards are empty) than the falsifying cause of death to pad numbers.

Which, to be honest, I was expecting purely based on how far out NY and NJ's numbers were from even cities-vs-state. I just didn't expect to be able to find any evidence.

I did know that the vast majority of risk was for people who aren't in the normal work demographics, in those cases where there wasn't padding.

It's just incredibly depressing.

Ohioan@Heart said...

Very depressing indeed.

Anyone caught padding or altering CoViD numbers, in any direction, for any reason (since these are to a federal administration and would affect their subsequent actions) should be prosecuted for making false statements under the federal statute.

Foxfier said...

Amen. All it takes is just stating what the bleepin' standard you're using is, and it's fine....

As was pointed out over at The American Catholic, its really hard to prove to the level needed for action. :/