Read on Twitter
There were to be 125 million Covid infections in India by mid-May, according to some estimations. What happened?
Mathematical models, esp. when coming from big scientists are usually taken at face-value by the general public. They are not meant to be. Those early models? Their utility was limited to "it's gonna get bad -- be prepared", and that's according to the people that made them
So is that all models are useful for? Certainly not. Models coming out now can be incredibly useful for planning post-lockdown measures. A model by IISc-TIFR, for example, examined how trains can resumed in Mumbai and suggested that they operate with 20% capacity.
They did this by simulating the *entire city* and seeing possible outcomes. A video game. A data and computationally intensive process. To do something like this, you need a hell lot of data, both before and during the pandemic. Something India has largely not been doing well.
The MoHFW, in fact, only compiles data related to the number of cases--active, recovered, deaths and the likes. Granular data about local outbreaks, reporting time, recovery duration, etc is either not available, or available for some cities and not others. Modeller's nightmare.
Dr Giridhar, Lifecourse Epidemiology, PHFI told me about how govt meetings work
- "Each one will be saying this model says this, that model says that. Nobody will acknowledge which model finally [was most useful]. They will have a different way of analysing this inference."
- "Each one will be saying this model says this, that model says that. Nobody will acknowledge which model finally [was most useful]. They will have a different way of analysing this inference."
And that's another problem. We don't have a culture of collaborating with scientists. Work has largely been in silos. We need to sit down with our best minds and work *with them*.
And that's going to be doubly important with lockdown restrictions easing. Local models are going to be at the forefront of planning. And that's my story in today's @TheKenWeb. Free to read -- do share widely. https://the-ken.com/story/indias-next-top-covid-model/
