There are a whole bunch of 2019-novel-coronavirus related notebooks, and this is the kind of event with large-scale public interest where data visualization and mathematical modeling is at its most powerful to influence people’s behavior and potentially save lives.
It would be neat if the Observable team could curate a collection of the best relevant notebooks (ordered by usefulness/relevance, not by creation date, so that the best stuff is most visible), put it at the top of the front page, and try to encourage the community to work with epidemiologists, virologists, doctors, et al. to (a) effectively communicate to the public what is happening, what is likely to happen, and what people can do about it and (b) help put together models where parameters and assumptions are tweakable, so experts can have a better tool than spreadsheets or shell scripts or whatever they are currently using to share ideas.
One of the best epidemiology summaries I have seen is a Medium blog post written by a Bay Area marketing executive, but it’s all statically rendered charts and models. Observable is a potentially better platform for this kind of thing if the plots can be rendered at runtime with the modeling done in the page, so that changing parameters is accessible to readers.
We’ll think about how we can surface this collection. (We might also do something search-based so that any notebook can be part of the “collection”, and then use likes to sort… but we’ll see.)
I agree with Jacob on the second part of his message too: “encourage the community to work with epidemiologists, virologists, doctors, et al.”.
Probably this ought to be discussed and carried by the DVS, but what we (observable users) could do is collaborate on some notebooks; we could probably create an organisation to make that collaboration more fluid—especially if we don’t want to publish unfinished/unvetted visualisations.
One scary thing is that in the USA we are doing dramatically less testing than Italy, so if anything we are closer to their situation than it currently appears. Likewise, South Korea did much more testing than any of the other places listed, so their numbers are going to overestimate their current situation relative to others. I don’t know if there’s any statistically valid way to compensate for that though.
Yeah. We had to fudge the numbers for Boston a bit due to the dramatic undertesting. As of now, they are undertesting at 10x relative to other states in the US…
My 2cts after discussing with a few epidemiologists.
Testing will yield numbers proportional to the population being tested, so those numbers of “confirmed cases” are highly sensitive to test availability.
A different way to count is to base your denominator on a certain condition that is sure to be detected. Conditions that require hospitalization in ICU are much more similar from one country to the next. You can then reverse-compute the incidence of contamination by making an assumption on that ICU rate wrt infected population. (It’s of course much more difficult, because you need to factor in the time it takes for one person to develop the disease, and the huge variation between age groups and (probably) other social conditions.)
Still, to compare countries (or countries trajectories), it makes more sense to compare the number of deaths than the number of positively tested cases. But this is still a very rough estimate.
In terms of forecasting, epidemiologists apply several rules of thumb that work, more or less, for the flu (which doesn’t like hotter weather, for example), but this disease is quite different, so these can go only so far.
Hopefully nobody minds if I keep linking to data graphics / graphical expositions from around the web. These from Simon Scarr and Samuel Granados at Reuters are nice:
I made a choropleth for county-by-county case counts in Rhode Island, as well as one that shows the percent of the population that’s infected, which helps balance out the large population differences between counties:
Unfortunately the RI government isn’t the greatest at publishing the data consistently so the county-by-county numbers don’t line up with the statewide numbers.
Relatedly, this is a good template for the county-by-county map/projection for RI in case anyone needs one in the future!
@yurivish - I am pleased that you changed the title of this notebook. It’s a great piece of work, but when I went to share with my colleagues under the old title, I couldn’t help but wince at the negative tone (though I totally got your point). The positive call to action is much better! Nice URL redirection too.
Yay! I really didn’t like that bubble chart so now I can make a choropleth!
(I would also like to do one that takes in county populations to determine % infected as I think that would be more useful than this, which looks a little like a population density map)
