Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Fine, I'll run a regression analysis but it won't make you happy

natesilver.net

Post reply on HN
Print view View original

211–220 of 341 posts

Re: Fine, I'll run a regression analysis but it won't make you happy

#211

Post by gdevenyi »

Aachen wrote: post #91
Amorymeltzer wrote: post #43

For those curious (as I was) what he's using, it looks like Stata: https://www.reed.edu/psychology/stata/analyses/parametric/Re...

Was looking for a download, but there's precious little about that on the website. The closest I got were these remarks: > On all Reed lab computers, Stata is located in the Applications folder. > Stata can be found in the Applications folder of any school machine. On a PC, it is most likely located in Program Files. Does that imply it's neither open source nor even commercially available? Seems rather odd, surely it…

Sata is proprietary.

Re: Fine, I'll run a regression analysis but it won't make you happy

#212

Post by bryanlarsen »

gpt5 wrote: post #201
slv77 wrote: post #172

Earlier quoted context omitted.

The cost-benefit calculations in terms of excess deaths are irrelevant. It was clear even during the initial outbreak that policymakers primary constraints were the due to uncertainty about the virus and limits on availability of protective gear and ICU beds. Without protective gear there was no way to reduce transmission without isolation and once ICU beds are at capacity mortality rates double and people are dying…

Initial constraints like ICU bed shortages and gear scarcity definitely had a basis. But those measure lasted long after the bottlenecks got resolved and the risk assessment became clearer. For example, public schools stayed remote while virtually every private school switched to in-person (even California's governor opted for private in-person schooling for their kids). The public's frustration is that these prolong…

Hospitals were closer to collapse in 2022 than they were in 2020 or 2021 in my experience.

Re: Fine, I'll run a regression analysis but it won't make you happy

#213

Post by defrost »

logicchains wrote: post #137
Waterluvian wrote: post #131

It was a hard thing to broach at the time without people frothing at the mouth about how I must be anti vaccine or something: Has there been any attempt to calculate the cost-benefit of all the various measures and how extreme we should go with them? I’m guessing it’s hard to quantify and compare. A lot of things like general depression, isolation, kids missing half a year of school, etc. can’t really be evaluated ag…

Yes, there are papers attempting that, such as this recent one, which looked at whether shelter in place policies reduced overall excess mortality: https://onlinelibrary.wiley.com/doi/full/10.1002/hec.4737 . "Using an event study approach and data from 43 countries and all U.S. states, we measure changes in excess deaths following the implementation of COVID-19 shelter-in-place (SIP) policies. We do not find that cou…

FWiW, because it's a very unique situation globally, W.Australia a state 3x the size of Texas with a population a bit over 2 million people effectively "sheltered in place" by airlocking its borders - international and inter state flights had to quarentine, road transport on the one major transport road to the eastern state was pick up | drop off with out of state drivers quarentined, shipping through the port was non contact with crews quarentined.

W.Australia's COVID-19 vaccine roll-out began 22 February 2021.

The state remained isolated from 24th March 2020 until 3rd March 2022, by which time essentially every person in the state had been double vaccinated.

For the most part people continued life as before, now and again the capital city would have portions "locked down" when there was an active case in order to contain spread and test potential carriers.

The first active case was the 21st February 2020, contained. The first COVID death was 1st Mar 2020. Active case numbers stayed low until There was no real increase in active cases until February 2022. There was no significant increase in deaths due to COVID until March 2022.

"Shelter in state until vaccinations rolled out" for a state that was already largely independant and isolated worked just fine - the deaths started once the borders opened, but at rates greatly below those seen elsewhere in the world thanks to a vaccination buffering.

W.Australian cases Jan 2020 - now: https://covidlive.com.au/report/daily-cases/wa#2020

W.Australian deaths Jan 2020 - now: https://covidlive.com.au/report/daily-deaths/wa#2020

https://en.wikipedia.org/wiki/COVID-19_pandemic_in_Western_A...

Re: Fine, I'll run a regression analysis but it won't make you happy

#214

Post by dalmo3 »

Waterluvian wrote: post #183
ryandrake wrote: post #176

Earlier quoted context omitted.

Almost nowhere on earth did "stay at home" actually happen. We all had these so-called mandates, but people (especially in more rural areas) pretty much did whatever the hell they wanted with no repercussions since there was almost zero enforcement. They were out and about, shopping for their khakis at "essential" businesses, eating at the restaurants who refused to closed (also with no consequence) and basically ign…

You clearly didn’t live where I did. Fascinating how natural it comes to us: “my experience must be everyone’s experience.”

Also fascinating how the multiple meanings of the word "must" apply to your quote.

Re: Fine, I'll run a regression analysis but it won't make you happy

#215

Post by dghlsakjg »

VancouverMan wrote: post #145
rcbdev wrote: post #126

I don't know about other countries but in Austria nurses were highly incentivized to report deaths immediately following a vaccination or weeks after an infection has passed as a covid death in the sense of national statistics. I would not be surprised if this "nudging" of on the ground reports over the course of the pandemic has rendered the data around covid deaths unreliable.

The death counts from Canada from that era are very questionable. As early as June of 2020, the "public health" authority in Canada's most-populous city (4th highest population in North America) revealed that the counting of deaths was being done using a methodology that sounds quite dubious: "Individuals who have died with COVID-19, but not as a result of COVID-19 are included in the case counts for COVID-19 deaths…

Attributing what happened in Ontario to the whole of Canada is something that happens way too often.

As a Western Canadian resident, I don't think it's fair or reasonable to extrapolate about all of Canada from a single tweet from a TO health authority, and an article with a disclaimer about it's own accuracy.

There well may be problems, but please keep in mind that while 50% of Canadians live between Montreal and Toronto, 50% of Canadians don't, and many of us live lifestyles that have absolutely nothing to do with that region and it's governance.

Re: Fine, I'll run a regression analysis but it won't make you happy

#216

Post by slv77 »

gpt5 wrote: post #201
slv77 wrote: post #172

Earlier quoted context omitted.

The cost-benefit calculations in terms of excess deaths are irrelevant. It was clear even during the initial outbreak that policymakers primary constraints were the due to uncertainty about the virus and limits on availability of protective gear and ICU beds. Without protective gear there was no way to reduce transmission without isolation and once ICU beds are at capacity mortality rates double and people are dying…

Initial constraints like ICU bed shortages and gear scarcity definitely had a basis. But those measure lasted long after the bottlenecks got resolved and the risk assessment became clearer. For example, public schools stayed remote while virtually every private school switched to in-person (even California's governor opted for private in-person schooling for their kids). The public's frustration is that these prolong…

ICU capacity was a bottleneck that never got resolved. In the US there was never a nationwide effort to mobilize ICU resources to target hot spots other than an initial aborted attempt to gather respirators for the surge in NYC and send a military floating hospital.

Regional programs were put into place to shift patients during a local surge and to mobilize hallways and other non-traditional capacity. In California once capacity dropped below 10% health orders went into effect.

At the beginning of 2023 and well after widespread vaccination the California statewide capacity was at 24% availability with 7% of beds being taken by Covid patients.

https://www.cdph.ca.gov/Programs/CID/DCDC/Pages/COVID-19/Reg...

Re: Fine, I'll run a regression analysis but it won't make you happy

#217

Post by jalapenos »

Waterluvian wrote: post #131

It was a hard thing to broach at the time without people frothing at the mouth about how I must be anti vaccine or something: Has there been any attempt to calculate the cost-benefit of all the various measures and how extreme we should go with them? I’m guessing it’s hard to quantify and compare. A lot of things like general depression, isolation, kids missing half a year of school, etc. can’t really be evaluated ag…

This utilitarian reduce-to-numbers approach misses the important matter of the inevitability of death.

In the grand scheme of things, all those who died purely from COVID were going to die anyway.

A more important question is what has happened as a result of societies responses to this pandemic, especially in light of how marginal this virus' lethality was (this was no plague).

It has set precedent that in the case of relatively minor social distress, the government can take unlimited power, and the populace will just knod their heads.

It has also set precedent that the old can parasitize the young even further. Saving a few boomers a few years is worth masking up young school kids, affecting their communicative and social development.

The only silver lining to this has been the resulting explosion of remote work, which of course the boomers are now trying to undo. Fortunately, that's a cat-out-of-the-bag.

It's unfortunate that the boomers didn't learn an important lesson from their parents, who fought WW2 - sometimes dying is the right and proud thing to do.

Re: Fine, I'll run a regression analysis but it won't make you happy

#218

Post by slv77 »

somsak2 wrote: post #193
slv77 wrote: post #172

Earlier quoted context omitted.

The cost-benefit calculations in terms of excess deaths are irrelevant. It was clear even during the initial outbreak that policymakers primary constraints were the due to uncertainty about the virus and limits on availability of protective gear and ICU beds. Without protective gear there was no way to reduce transmission without isolation and once ICU beds are at capacity mortality rates double and people are dying…

It's not irrelevant, especially considering there were restrictions lasting far longer than "initial uncertainty." It's not like your only choice is to allow for a "collapse" -- as an extreme example, you could decide you're not going to treat covid at hospitals at all (by the way, we make such determinations for other conditions or diseases all the time).

ICU capacity consists of the room, the equipment and the staffing. Once oxygen saturation starts to drop the treatment options are oxygen and eventually intubation and a medically induced coma. There is no realistic scenario where untrained or semi-trained individuals are going to be able to provide that support and even if the staffing shortages could be worked around there wasn’t adequate equipment.

The at home treatment option was to watch the patient drown in their own fluids. There are few people in that scenario that would have the emotional capacity to calmly give support to a close family member or friend when there was no ICU capacity. Their more realistic scenario would be people loading the patient into the car and driving to the nearest hospital. Some percentage of that population will likely turn violent.

Re: Fine, I'll run a regression analysis but it won't make you happy

#219

Post by xyzzy_plugh »

robhunter wrote: post #173
brody_hamer wrote: post #141

Earlier quoted context omitted.

In Canada (and likely elsewhere), there were a number of restrictions which were tied to hospital capacity. As the number of available ICU beds dropped: more restrictions were put in place, and vice versa. A lot of emphasis was placed on preventable deaths. That is; situations where we’re turning away a patient that we could’ve saved , because we don’t have a bed available for them. I thought this was a solid, intuit…

Fellow Canadian here - I think the suggestion that restrictions were tied to hospital capacity is a bit off. Restrictions were largely political - driven extensively by polling - and changed frequently with minimal or nonexistent reasoning. Here in Ontario, the McKinsey-driven "Science Table" frequently referenced hospital capacity in their reports, but their modelling often ended up being off by orders of magnitude.…

How odd. You must live in a different Canada than I do.

Re: Fine, I'll run a regression analysis but it won't make you happy

#220

Post by lazide »

panarky wrote: post #206
Blackthorn wrote: post #189

Earlier quoted context omitted.

No, they're saying that excess deaths is a rearward looking metric that nobody had the luxury of actually having at the present time when decisions were being made. It's very easy to make decisions in retrospect when you have all the information that could be useful. It's very hard to make decisions in the present when that information doesn't exist yet.

>>> What about ten lives? One thousand? Ten thousand? Are 200,000 lives meaningful enough? Silver's regression model shows that for each 1% increase in the vote share for the blue candidate, the covid death rate in that state (statistically) would fall by 15.5 per million population. In areas where the blue candidate received 20% of the vote, and the red candidate received 80%, the model predicts 1,793 deaths per mil…

Arguably, ideology is and has always been the largest contributor to death (and birth) rates no?
Post reply on HN