Forecasting for COVID-19 has failed

[PeterNSteinmetz]

Commentary by Ioannidis at Stanford on the lack of accuracy in Covid-19 models.

Since we have a lot of other Covid-19 threads, I would suggest this one be used for discussions of the models and the accuracy/inaccuracy.

https://forecasters.org/blog/2020/06/14/forecasting-for-covid-19-has-failed/

 

[PeterNSteinmetz]

I'm surprised that people need to be told that the models would not be perfect/exact and that some may be wrong. People really thought we had 100% understanding of a virus that the world had never seen within months?

Sometimes of course the model outputs were implicitly embedded in other things people were told.

For example, most of the claims which politicians have made that their interventions reduced Covid-19 cases or deaths were based on comparing a model output of higher cases and deaths than observed and a model prediction of higher cases or deaths if an intervention term was dropped. That is different than a statistical comparison of two observed cases, one with and one without the intervention. But most people likely did not understand that the claimed advantage was just a model output and not a statistical comparison.

 

[PeterNSteinmetz]

Unfortunately, there are in fact a lot of people that act like M&Ms are the end of the world.

You mean, not having an adequate supply is the end of the world :emoji846:

 

[PeterNSteinmetz]

We’ll see how far this goes. At least in AZ, the rate of growth of new cases actually trending down since about Memorial Day, which seemed to cause a huge burst.

 

[PeterNSteinmetz]

Not exactly. I saw with my own eyes an official at the podium explain that the Covid-19 death tally included all people who died with Covid, not necessarily those whose cause of death was Covid. That's fine, as long as we all know what that statistic means and doesn't mean.

There is always some arbitrariness in assigning the cause of death, especially in the case of deaths which have multiple health issues. Agreed it would be nice if there was a a standard on this.