4 ms·
I'm one of the researchers on a new disease forecasting project, and I'm bookmarking this because we often use hurricane forecasting as an example of a level of
by Fomite 3y ago
I'm one of the researchers on a new disease forecasting project, and I'm bookmarking this because we often use hurricane forecasting as an example of a level of accuracy we aspire to, so this is a really interesting breakdown.
- uticus 3y agoWhat forecasting models / techniques are prevalent in disease forecasting?
- Fomite 3y agoThere's usually a mix of statistical models (some traditional time series methods, some neural network based) and mathematical models (i.e. SIR models). None of them fully capture all the dynamics we need to.
- KRAKRISMOTT 3y agoDifferential equations make the world go round.
- HPsquared 3y agoThat, and conservation laws. (Angular momentum)
- jacquesm 3y agoThe big miss here seems to be that not enough data had been seen yet to even know the envelope of what's possible. Without that there will always be inputs where a model breaks down. Disease modeling is also always going to have to contend with the fact that our historical data was collected when things weren't quite what they are today, for instance population numbers and densities are still entering uncharted territory compared to the past.
- ethbr1 3y agoIf author's hunch was correct, it may have been a "greater than the sum of its parts" hyperlocal edge case. E.g. the wind sheer exactly right, a patch of warm water, etc. That does seem really tricky to model without making the models overly jumpy.