Hacker News Reader

Comment by cmrivers | Hacker News Reader

The outbreak in Uganda is troubling, but a forecast that goes out 6+ months is silly. No professional infectious disease modeler will forecast out more than a few weeks because the trajectory of the outbreak changes in response to interventions and behaviors.

View on HN

>a forecast that goes out 6+ months is silly.

Sure is, which is why it's odd we've been bombarded with endless graphs with months of predictions, over the past 2.5 years.

Doesn't seem odd for news outlets, unfortunately.

Hyped up drama sells newspapers. Unfortunately.

For ebola?

How've I not noticed this? I've instead seen a whole lot of graphs with no prediction whatsoever and lots and lots of people in positions of authority who seem totally unable to extrapolate from those in their heads, resulting in idiotic back-and-forth on various restrictions for the first 18 or so months of the pandemic.

Shit like schools releasing plans ahead of the school year that they then immediately ignore because otherwise they'd have to close in the first two weeks of school, when it was fucking obvious the numbers would be like that around that time, just from looking at the graph and knowing more-or-less how disease spreads. Or "Ok stop masking and open up restaurants wait oh shit it's going up again I thought the tiny dip we saw would continue forever, for no good reason". Just baffling levels of data-illiteracy.

But not a lot of long-term prediction graphs. Who was publishing those?

[EDIT] Wait, I did see total-deaths-at-time-X predictions with/without measures, and with/without vaccination at high rates. That's true.

[EDIT EDIT] Is there a tone issue or is my having seen vanishingly few graphs for all of COVID that tried to predict trends more than a week or two out an outlier experience, and those were in fact extremely common in, perhaps, media I didn't look at? Truly, the main problem I saw locally was an astonishing near-complete failure to consider trends and likely projections, over and over again and often by the same people, who seemed weirdly incapable of learning a very clear lesson, rather than too many projections looking too far out.

My guess is a tone issue. It’s hard to even tell that you’re saying you haven’t seen long term graphs - the first sentence comes across as sarcastic when followed by long sentences complaining about other problems.

I thought:

> I've instead seen a whole lot of graphs with no prediction whatsoever

made it pretty clear, but maybe not. And the rest was expressing that the actual on-the-ground problem I saw, and the single biggest problem with my state & local-level response to the whole thing, was a complete lack of attention to future trends, not too much. But perhaps that doesn't come across very well. Mea Culpa.

Sounds like you're in the UK. Do you consume any mainstream media?

https://www.google.com/search?q=uk+covid+predictions&tbm=isc...

I didn't take in much news, and now avoid it more than ever, but what little I did was wall-to-wall with projections complete with big scary peaks and steep rises in numbers. I dread to think of the state of mind of someone who watched more like the average number of TV hours (for me it is zero), and took in all of this with an uncritical mind.

We have? I’ve certainly been bombarded with graphs showing trends, but six month predictions? Where are you seeing these? Can you point me to some?

Sure.

https://www.google.com/search?q=uk+covid+predictions&tbm=isc...

There are plenty that stretch out for a year.

A naive google search is ... not a helpful reference.

Why not? It's the most efficient way I could think of, to display the sheer quantity, timeline, and content of the "Covid predictions" that we were shown in the UK. They are easily clickable, to show the original context (typically a news story or press release). You'll also note that practically all of them extend for at least 6 months.

Also, you "asked for some". There they are - there's a lot of them.

Well to be frank I clicked a handful of them and got tired of trying to decipher graphs that didn’t end up supporting this claim.

Reminding ourselves that you asked to be "pointed to some (graphs predicting more than 6 months)":

It is not clear when the image in first result was published, but given that it includes error bars for dates from March 2020 onwards, I would assume this is the result of modelling performed earlier than March 2020. It covers until October 2020.

Result number two is a direct link to this paper:

https://www.imperial.ac.uk/news/196234/covid-19-imperial-res...

It was published in March 2020, and predicts data till March 2021

Result number three is a direct link to this news story (with an ironic headline, given result number 4 below):

https://www.bbc.co.uk/news/health-54831334

It was published in November 2020, and predicts data until July 2021

Result number four is a direct link to this news story:

https://www.bbc.co.uk/news/health-53392148

It was published in July 2020, and predicts data until July 2021

This is the garbage that was fed to the UK populace on a daily basis - it was like the "modellers" had taken charge.

Will you accept that graphs predicting more than 6 months of results were in common circulation?

> Will you accept that graphs predicting more than 6 months of results were in common circulation?

I accept that they exist. I don't accept that we were "bombarded" with them. This is, of course, a US perspective, so perhaps your mileage may vary.

I guess as it transformed from epidemic to pandemic to endemic, it became increasingly predictable. The duration and peaks of the last waves could even be predicted by months indeed, as measures and behaviors remained practically unchanged. Not one of these predictions encompassed six months, though.

>Not one of these predictions encompassed six months, though.

I beg to differ. Here's a sample of what we got, on a daily basis, in the UK:

https://www.google.com/search?q=uk+covid+predictions&tbm=isc...

Plenty of those are dated from the early months of 2020.

I stand corrected if that was the case. I followed WHO and Mexican predictions, and more speculative ones if the authors were honest about the speculation.

That sounds... wrong? What if the 6+ month forecast assuming no interventions is useful to compel an intervention, and the 2-3 week one is not?

Then it's a scenario model, which asks what-if questions usually for planning purposes (e.g., what's the most hospital beds we might need?). Scenario models can have longer time horizons than forecasts, but 6 months is still quite long.

That is a quite cool way of separating those models. Can you point me in a direction where I could learn more? RE: Forecast vs Scenario?

It's the same thing. Just the latter implies you're identifying an assumption or two as variable and modeling outcomes of some of them changing. They're all forecasts. To argue otherwise is semantic pedantry.

If “useful to compel” is a criteria, we can jump straight to flat-out lying and dispose with the statistics part.

The GP is invoking the idea of a compelling understanding, one that produces action by buy-in.

If you're confused on that point vs sinister authoritarian actions, then the risk of other people lying to you is far from your biggest problem.

So you're in charge of making decisions. Hypothetically. Your epidemiologists believe that if no action is taken, 10 people will die in the next two weeks, 100 people will die in the next 4 weeks, and 1,000,000 will die in the next 6 months.

Which forecasts would you prefer they keep to themselves?

Starting an implication with “if no action is taken after 100,000 deaths” is the same as “if false” — you can put anything in the “then” and have the statement be logically sound - that doesn’t mean it’s in any way useful.

End of the day that kind of talk just damages trust in the scientific process, which has to be at an all time low with regard to epidemiology, as a result of the stream of certified lies that have been presented as public policy over the past few years. (“Masks don’t work” being the first of many)

> Starting an implication with “if no action is taken after 100,000 deaths” is the same as “if false” — you can put anything in the “then” and have the statement be logically sound - that doesn’t mean it’s in any way useful.

1) Not when YOU are the person who determines if action should be taken!

2) You seem to be implying that action can be taken at 100,000 which will be effective at preventing the spread to 1,000,000 which may be true or may be false, but it's going to be far easier to stop it at 10 regardless.

> “if no action is taken after 100,000 deaths” is the same as “if false”

Is useful as the higher bound and for judging the results of actions.

But still... I think you're too sure about people in charge following reasonable behaviours. We don't even have to go as far as the great famine. There's lots of things going wrong now where people responsible explicitly do not address the issue.

We already did quite a bit of lying during the Covid pandemic.

Update, this comment refers to the original blog post submission, which pulled out forecasted numbers with no context. The Telegraph article is better.

How is it disingenuous to use assumptions for modeling?

That doesn't make sense. I'm no epidemiologist, but have friends who work in disease modeling, and used to follow some virology blogs/podcasts. They usually make lots of models, each with different assumptions, specifically to guess what the intervention of different public health measures would do.

How far out did the Covid forecasts look, to justify global lockdowns?

The fear with COVID was mutation of the disease into something more deadly and more contagious. What looks like fearmongering now was cautious behavior around mutation at the time. Conversely, Ebola isn't likely to mutate now, so we don't need that same level of caution.

> What looks like fearmongering now

We had over 1 million deaths in the US.

You travel back to 2016 and tell people we'll have a pandemic in the US, 1 million people will die, and then large groups of people will attempt to shrug that off as "nothing to get concerned about", people would find that incomprehensible.

If I were to post a comment on HN in 2016 suggesting that people would shrug off 1 million deaths due to a pandemic because it doesn't align with their worldview I would have been downvoted into oblivious for holding such a laughably ridiculous and cynical opinion.

Same could be said of the countless other climate related impacts that people in the very near distant past would consider horrific and impossible.

The power of the "this is fine" culture to not see reality is terrifying.

300k Americans die each year from obesity. Yet nobody seems to view this as an urgent issue needing addressing.

quite a lot of people do and have seen obesity as an urgent issue needing addressing. People at many levels of government, in both political parties, and for many years....

> Conversely, Ebola isn't likely to mutate now

Can you explain why?

They didn't need much more than noting Rt and experience of Wuhan, if you're thinking "IMHE!", sort of a red herring: yeah, they predicted X million of deaths, but it wasn't anymore complicated than take Rt, use in formula, and yeah we should probably see if we can eradicate this thing locally, early. Was never a global decision.

Are you claiming that the lockdowns that happened in the first half of 2020 had zero effect on SARS-CoV-2 infections?

They did not.

Florida vs California.

read up.

In case anyone would like to actually read up on the impact of restrictions:

States That Imposed Few Restrictions Now Have the Worst Outbreaks (Nov 18, 2020)

https://www.nytimes.com/interactive/2020/11/18/us/covid-stat...

I mean...

> Nov 18, 2020

Looks at deaths per capita, from Jan 2020 to today. The position of peaks is barely relevant compared to their cumulative impact.

I'm literally asking a question, since it is news to me that "no professional infectious disease modeler will forecast out more than a few weeks."

Do I sense attitude? If so, why?

That's not quite the right way of phrasing that question. Models are made with and without interventions taken into account, and the lockdowns were a reaction to models showing reduced deaths and hospital strain if lockdown measures were in place. I'm not sure what that guy is talking about.

But if you're actually interested, you should go and read the papers used for modeling yourself, since the odds of finding a rando on hacker news who is an actual epidemiologists is zilch. The ones who were public communicators that I followed over COVID were the podcast "This Week in Virology" which is just about the best resource you're going to find, and the blog "your local epidemiologist", who is also fantastic

Reply on news.ycombinator.com