Select countries by increasing / decreasing values

[andrewmackie]

Users are presumably interested in seeing which regions are currently 'winning' or 'losing' with respect to covid.

I have, therefore, added buttons and methods which will select all of the regions which are going up or down (with respect to the selectedData).

This is done by checking whether the sum of the slope over that last 4 days is more or less than the sum of the slope in the previous 4 days (where the 4 days is stored in the 'daysOfIncrease' variable).

If you adopt this change, it may be desirable to allow users to change the daysOfIncrease variable (as can currently done with doublingTime). I'm happy to make this change (or others) as might be useful.

Thanks for covidtrends - I appreciate it.

[netlify]

Deploy preview for frosty-benz-8c81bf ready!

Built without sensitive environment variables with commit 91d1d76

https://deploy-preview-209--frosty-benz-8c81bf.netlify.app

[rpkoller]

hmmmm a few thoughts about your pull request. At first by selecting the decreasing and increasing countries on a country level you get a not really readable and comprehensive overcrowded graph:

"comparisons" might be difficult in-between countries that way ( you would have to manually deselect x number of countries to narrow down the selection of particular increasing or decreasing countries - which implies maaaanyclicks) and you also have no real visual as well as readable feedback what you have actually selected and seeing right now (was it increasing or decreasing i had selected).

on a regional level it is a little bit more accessible due to the lower number of regions per available country. but it is still a visual overload and in case of zero cases (decreasing) e.g. for china the curves are missing many of the labels cause of the zero cases issue with covidtrends:

and about the decision if a country has increasing or decreasing numbers based on the last 4 days... isnt that interval too short and might give false impressions in each of the direction it might be read? would it be more reasonable to choose at least the 7 days we have for the running average for new confirmed cases on the y axis for normalization? but for the determination of the general trend i would suggest an even higher value? 14 days or even 28 days? so you are able to see that there is a significant overall trend. an example, in case of germany we had a decreasing growth overhere ...

then they started to open up the country step wise which lead to small spikes up for a few days but the general trend was still falling/decreasing.

a few weeks ago superspreader events started at slaughterhouses which lead to a more consistent uprising exponential trend alongside more and more people acting like the pandemic with no social distancing and so forth ( they did that during the time of the small spike as well already):

so a too short interval for the decision if a curve is increasing or decreasing might support and nourish that behavior even more. meaning people take a look at a curve and see hey my country is doing fine according to covidtrends and things are decreasing so i can out and "party" and the other way around the curve is increasing i better barricade myself at home ;) exaggerated but especially the former could be the way the feature could be read and used in arguments. ;)

and if the feature is merged i would definitely document in the documentation as well as the website copy on what basis the decreasing/increasing is calculated and the line is drawn in between the two groups. and why only use the two? an idea might be the following, @aatishb once linked an article where they used 4 categories to explain a curve based on covidtrends: exponential growth(steadily rising curve), linear growth ( more or less horizontal curve) , decreasing growth ( decreasing curve) and contained (more or less vertical curve - china is a good example).

[andrewmackie]

Thanks @rpkoller for your feedback. I'm very much open to an alternate or more comprehensive approach to determining whether curves are going up or down. Given the number of people using this site it probably needs a more methodical approach than the one I have suggested. FWIW I chose 4 days because longer time frames (especially out at 7 days) gave false positives / negatives.

I do believe that users would benefit from tools which allow them to quickly see which regions are doing well and which are doing poorly, particularly with each state of the USA, because this does inform policy and behaviour.