Add vignette for sircovid

[johnlees]

Still needed:

• Get a decent fit (I think just a problem with the MCMC? Or data and parameter spec?)
• Check if there is an automated way to look at the user parameters needed for an odin model
• Add some plots of the results
• Look at the Rt & plot it
• Add a section on forecasting

[codecov]

Codecov Report

Merging #137 (d7ce337) into master (07fb467) will not change coverage.
The diff coverage is n/a.

@@            Coverage Diff            @@
##            master      #137   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           10        10           
  Lines          702       702           
=========================================
  Hits           702       702           

Impacted Files	Coverage Δ
R/carehomes.R	100.00% <ø> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 07fb467...d7ce337. Read the comment docs.

[johnlees]

This should be relatively quick to run, but copy of output included here too:
vignette.zip

[johnlees]

Would appreciate comment on whether the saveRDS/loadRDS used for the long runs here is viable. Also the longest one is ~80Mb, so whether I should make a greater effort to prune that down (though I'm not immediately sure how)

[richfitz]

We will arrange some other way of orchestrating the long runs, probably along the lines of redux's offline vignette approach. Once people are happy with the science side I'll look at this, then squash out or filter the large objects

[richfitz]

If these are publically available, I'd rather explore downloading on demand than including them, as they will be forever out of date

[johnlees]

I've given links to all of these datasets, both the landing page and the specific version downloaded. I think we should keep these specific versions because the serosurveys are cumulative and don't have the time series (hence the inclusion of two to get timepoints), and because the data used is specific to certain dates, as noted in the vignette. So we don't want to update any of them.

[richfitz]

We will arrange some other way of orchestrating the long runs, probably along the lines of redux's offline vignette approach. Once people are happy with the science side I'll look at this, then squash out or filter the large objects

[johnlees]

TODO:

• Update to use coef() (when supported in dust and odin.dust)
• Fix generation of large objects