Rework graphic layout. Implement negative case method. Add notebook for non-US data

[pati610]

Hi Will,

I've spent a lot of time digging into your model, and intuitively – as well as logically, and visually – it strikes me as being very sound. I've reworked the graph layout to resemble the rt.live theme for easier comparison(Stan R0 US.ipynb), and I've also made a country-based notebook (Stan R0 EUW.ipynb), for now fetching mainly EU data but adjustable to include any country. I deliberately left your and Kevin's notebooks untouched.

The other modifications I've implemented are listed in the readme file in my repo.

I've put high res png files of the last runs in my image folder. It makes for some very interesting reading. One thing that strikes me in particular is how well your model lines up with the graphs at epiforecasts. There seems to be a slight skewness date-wise, but the curvatures are generally a close match.

Strangely enough, the graphs I get using the US mcmc model for the that was the prototype for rt.live are nowhere close to what's published on their web page, but I guess they've moved on without updating the git repo.

To add to the confusion: Running both the mcmc and Stan models on EU countries, your model is much more in line with their estimates.

Thanks for documenting your method so well, I learned a lot from the process.

I'm thinking about making a graphic layout in the epiforecast style as well, which probably would aid in the side-by-side comparison.