arshyam wrote:
I have been tracking Prof Agrawal's tweets (where he shares his graphs) and the fit is indeed remarkable. What gives hope is that multiple cities/states are now trending down, so that should ease some pressure off healthcare infra.
However, I wonder what this graph looked like in the February-early March timeframe? I looked around, but couldn't find any older charts. The reason for my Q is rather obvious, but will state it for clarity: did this model predict this steep spike in April? Last I recall, per this model and your earlier posts, this thing should have died out in February, but it didn't. I, for one, had set a lot in store by this prediction (my pandemic fatigue played a role too), but turned out to be so far off the mark. One wonders if the govt felt the same way too?
Prof Agrawal, and few others from the super modeling team (and also from DST and IIT's) are now a little more active on SM and those tweets are an excellent resource - The Indian Newspapers have done some *really* bad job in reporting in the past - IMO there should have been a good public domain data-sharing would have done better. I think a good site will be up soon. The parameter(s) (which changes over the time) and method itself is well published so one can do one's own calculation to see various scenarios. Had some interesting conversation with these professors, let me post my take (and stories which are already well known) wrt to your queries (and most media stories about "failed model" etc.
Some points: (This is *my* take - and my take only - take it FWIW) Sorry for long post ..
1- It is mathematical modeling - not absolute correct "prediction" (as in astrology), viruses don't behave exactly like your model and one learns more science..
2 - True: The "model" (initially IIRC in July or so)predicted peak in September and generally dying down around last week of February (even without vaccines).. No second wave (generally speaking) was expected then. This is what I posted here in brf too.
3 - True - the team expected mutation and even pandemic fatigue (people will be not wearing masks that religiously etc) but they were still fairly sure of no serious "second wave" - some minor hiccups, local flare-ups etc but not what we have now.
4 - We expected mutation but *no* one could guess it will be that bad (B117, P1 and our own B1.116) with *such* a high R factor (beta parameter - in super model) - Since India does not have good gene sequencing we were late in finding how prevalent these VoC were in India. (The data from UK/US/Canada worried many of us tremendously - I literally remember getting very scarred just looking at the graphs - the situation will become 'horrendous' etc .. but were hoping it will not spread to India - More of this later - when more data is available). (I remember one person, looking at some Canada data with P1 variants - and literally remarking "it's the end of the world" if it spreads to India - was that scary)
5 - Data till the end of February was still matched - relatively good fit - the projection (death rates came down to 100/day or so etc)
6 - Some got worried that (around end of Feb) that the "decay" was not fast enough -- but still even in the first week of March many were "hoping" (or wanted believe it's just local variation - not change in parameter)
) that there will be no nation-wide "second wave" and places like Mumbai and Punjab will be contained. (Many remote places in India still do not show the second peak till now).
7 - By the end of March everyone was sure of the second wave, was real , its exponential growth, it's timing etc.
(I still have all the graphs (or parameter values) shared by Prof Agrawal - good learning material -- Basically model is quite reliable .. The "big" phase transition (which in model takes about 15 days to stabilize ) makes even best of us "wrong (mostly uncertain)" in their prediction. The model and data have been fitting pretty well for last few weeks.
(Just for perspective in the beginning of April one of the chart showed 170,000 (cases/per day), my calculation was closer to 250,000 cases/day - best we could say was the peak is around this time (much more certain) and height would be 170K (or 2 or 3 times that - depending on various factors) - the value is around 400,000 now)
Sorry for the long post -- May be brf is not a place for this but some parts take a long to explain. There are some very good resources about the model and data.. etc.