Some small suggestions

[wihlner-addcream]

Just a few observations.

1. Reporting delay The glanceability of the first chart is great. But the ones below it are a bit harder for me to understand. Possibly because I'm math'icapped. What I've been missing is a graph similar to the first one that shows death dates for the deaths that have been reported today only.

2. Navigate between dates It would be interesting if I could go back and see what the graphs looked like on a specific date.

3. The font weight of the graph headers seems to be set a bit too high This is clearly the most important point :)

[wpegden]

I agree with suggestion 2. That is what I came here to suggest!

Even just don't delete the old files from the web directory; then we can manually change the date in the filename for the png and see the old predictions.

[morberg]

Even just don't delete the old files from the web directory; then we can manually change the date in the filename for the png and see the old predictions.

The old graphs are available.

[morberg]

I've made an interactive graph where you can click and drag on the bottom graph (showing number of reported cases per day) which will filter the top graph showing actual death dates.

You probably want to select the leftmost bar and then drag to the right, but you can make other selections as well.

Edit: I've added an animated GIF showing the usage on https://github.com/morberg/covid-notebook if the instructions were unclear.

[morberg]

1. Reporting delay The glanceability of the first chart is great. But the ones below it are a bit harder for me to understand. Possibly because I'm math'icapped. What I've been missing is a graph similar to the first one that shows death dates for the deaths that have been reported today only.

I just realized that by making the selection just one day wide you can see just what you are after. If you want you can then drag the selection to see other reporting days as well.

[adamaltmejd]

Thanks for answering @morberg! Re 1: I agree. Haven't figured out a better way to do it though. The important thing is to achieve some way of tracking the size of the delay over time. So one can see if its changing, like it did over easter.

[morberg]

I made an attempt in a different issue comment, but it was a bit cluttered. This might be better?

If you plot each individual reporting day it gives an understanding for each day, but gives no sense of trend:

[albertogomcas]

I would also add a line like the 90 percentile but with the median.

[wederbrand]

@morberg Your interactive graph looks really good. I came hear to suggest something similar. What I'm really after is the historical forecast, if you can add the historical forcasts to your interactive graph I think it will be as perfect as can be!

With the forecast chaning each day we can get a grasp of how accurate it is, or isn't.

[morberg]

@wederbrand: Here you go.