Models can be useful for extracting insights from patchy data early in an outbreak, but once we have direct measurements that can answer a question, it doesn’t make sense to rely on modelling estimates.
Models help us get more out of data. They aren’t a substitute for it.
Models help us get more out of data. They aren’t a substitute for it.
As an example, we estimated 4-7% of Wuhan had been infected by end of Jan (thelancet.com/article/S1473-…). A later serological study in April found ~10% (onlinelibrary.wiley.com/doi/pdf/10.100…). So model was probably in right ballpark. But ideally we’d of course rather have data than model estimates.
Here are some other early COVID knowledge gaps that modelling helped address:
