Originally posted by photoptimist Yes, GIGO affects all models, especially forecasting models that combine exponential growth of the virus and inexplicable human behavior.
However, there are a wide range of computational, mathematical, statistical, and scientific approaches that can tackle the problem. They can make the most of the sparse data we do have, account for the uncertainties in that data, correct for likely biases, and forecast the likely range of outcomes. For example, some methods use different assumptions about the numbers to make different forecasts and then watch the outcomes to judge which assumptions were likely right or wrong. Moreover, those approaches can tell us what data is the most important to collect (hint: it's early and repeated testing to determine whether the disease is in the community, how fast it is spreading, who is getting it, who is spreading it, etc.).
As bad as models are, they beat the alternatives six ways from Sunday.
I'm just amazed at how quickly the IHME model is diverging from reality. I'm focused a bit on Virginia since I live here. By that model, there should be 97 deaths and 1051 hospital beds predicted as needed for April 6th. The state website says that currently there have been 54 deaths and 497 hospital stays as the result of COVID in Virginia (not sure if all of those patients are still in the hospital or not). So when the projection varies that much from the reality in just a few days of predicting, it tells me that the long term projections are going to be much worse.
It feels like weather forecasting, where short term forecasting, a few days from now, will be relatively accurate, but anything farther out is a crap shoot.
Maybe the biggest thing is that differences in verbiage in state mandates influence predictions. So, Virginia has not officially close non-essential businesses, but they are pretty much closed at this point. Certainly no more remain open than in states like Kentucky and Ohio where they have officially closed them.
Anyway, I suppose in the end it is probably best that people are scared to minding their distance in the short term if it means that we can get back to normal a little quicker, but it certainly doesn't make me feel good about our ability to predict the future, even in some limited sense.
Psychohistory this ain't...
(I guess we need Hari Seldon about now)...