epidemics/covid

Display peak and aggregate epidemic sizes on model pages

Closed this issue · 16 comments

At country pages, derive size of the peak for each trajectory - likely makes sense to take all trajectories, find peaks, and display as range

(At the peak of the epidemic, under this scenario, there will be 10M-30M active infections, requiring 1.1M to 3M hospital beds and 500k-1M ICU units.
Overall there will be 200-800M infections, which can lead do 2-15M deaths.)

Do you mean the map or the model pages (or the same)? (as the title is different from the description)

As the number stored in the json file are number of people infected per 1000 people, we would need to use population numbers.

@jk3141 The aggregate is simply the integral of the curves right?

yeah, it's the integral, but actually will be directly in some file as the cumulative bumber. and yes will need either multiplying by population or displying as percentage

This is a duplicate of #227

#277 got closed, so do we want to close this one too?

No, this should stay open! It's quite a high priority.

The thing that got closed was making it into graphs of stacked bars. Just displaying the total numbers is still a highly requested feature.

@jk3141

Overall there will be 200-800M infections, which can lead do 2-15M deaths.)

How do we compute deaths? Plain percentage (or percentage range) seems too simplistic, but would work.

How about a model with

  • One percentage of all
  • Additional percentage of people needing but not getting ICU
    • This further depends on how long they failed to get it / how long theu use it? Any simpler but still correct model with less assumptions?

That model seems good to me, as long as we add some error bars.

Hey. I finally moved into this, here are my thoughts:

"Algorithm"

You mention above that it should be an integral - but I don't understand why? As we simply have % of population who are sick, don't we just:

  1. find a day when all trajectories "peak" (e.g. their sum is max)
  2. compute statistics based on values of individual trajectories (min, max, std for error bars)

@hnykda I think the intent was to simply write the numbers in text below the graph. And there, we could take the range of the maximums from our traces. (Or something bit more complex, e.g. fit them with some distribution and display e.g. 90% CIs, or just discard top and bottom 1, showing this as 50% CIs ..)

I know - but what numbers would you write there? I don't think it's well defined thing, I can interpret it in few different ways (+ Jan mention some I don't understand). 😁 Is the one I described above reasonable?

Overall there will be 200-800M infections, which can lead do 2-15M deaths.)

I will add this to the JSON data (in ~30 minutes? working on something related ..).
Note: It may make sense to display this in % as well.
I will leave out the deaths (not sure about the formula)

At country pages, derive size of the peak for each trajectory - likely makes sense to take all trajectories, find peaks, and display as range

I think this is just the range of curve maxima. I can add this to the JSON too, if you wish.

(At the peak of the epidemic, under this scenario, there will be 10M-30M active infections, requiring 1.1M to 3M hospital beds and 500k-1M ICU units.

I am not sure how to compute this ..

I am not sure how would error bars for values of individual trajectories computed. What do you mean?

Also, I guess this all is per mitigation strength? @jk3141

(resolved with gavento from call)

#287 resolved part of it, there is a followup in #288