Add 'ready reckoners' vignette

[adamkucharski]

A useful tool during COVID was the 'ready reckoner' plot by Ellen Brooks-Pollock et al, which mapped social distancing scenarios to the reproduction number - see below image as example from this paper. By combining socialmixr data by setting with r_eff() from {finalsize}, this should be relatively straightforward to include as an example for users for basic social distancing (but some more thought would be required to incorporate assumptions about contact tracing etc, perhaps linking with {ringbp} development).

[pratikunterwegs]

Just chiming in because I've been using {finalsize} and would find this functionality useful as well.

IMO, a more general way to go would be to add a contacts scaling argument to r_eff() and final_size(). This could be as simple as passing a single number to scale the contact matrix, but looking at the example it seems it might be more useful to be able to scale the contact matrix for each demographic group as in {epidemics}.

This reprex shows (my understanding) of how it would be done:

# where 1.0 = no contact reduction
contact_scaling = c(1.0, 0.95, 0.93, 1.0, 1.0)

# scale both incoming and outgoing contacts, but prevent squared scaling
n_demo_grps <- length(contact_scaling)
contact_scaling_incoming <- matrix(contact_scaling, n_demo_grps, n_demo_grps)
contact_scaling_outgoing <- t(contact_scaling_incoming)
diag(contact_scaling_outgoing) <- 1.0

# assuming some 5x5 contact matrix, each element [i, j] is scaled by the
# corresponding element of this matrix
contact_scaling_incoming * contact_scaling_outgoing
#>      [,1]   [,2]   [,3] [,4] [,5]
#> [1,] 1.00 0.9500 0.9300 1.00 1.00
#> [2,] 0.95 0.9500 0.8835 0.95 0.95
#> [3,] 0.93 0.8835 0.9300 0.93 0.93
#> [4,] 1.00 0.9500 0.9300 1.00 1.00
#> [5,] 1.00 0.9500 0.9300 1.00 1.00

It would be possible to modify the contact matrix expansion here - repeated in r_eff() to include scaling of contacts by susceptibility group as well. For more fine-grained contact scaling though, one would need to pass a large matrix that gives the full scaling by demography-susceptibility combination (e.g. shielding elderly with low immunity) but perhaps too complicated for users?

Happy to volunteer this feature unless somebody is already working on it.

[adamkucharski]

That would be great – there's no active work on this yet. As a starting point, having fixed susceptibility and ability to output r_eff by different contact assumptions (e.g. a list of contact_scaling vectors or similar) would be nice – and hopefully simpler to implement for a V1.

[adamkucharski]

Also adding a link to the original code for the ready reckoners analysis by @ellen-is: https://github.com/ellen-is/reckoners - would be great to get her thoughts (perhaps once we have prototype).

[pratikunterwegs]

Thanks, what I can do is make a small PR with the contact scaling functionality added and a brief vignette showing similar plots. That wouldn't include any uncertainty as that's coming from bootstrapping which might be a bit much for a first pass.

There are quite a few scenarios in the repo where I haven't looked too closely at how they're translated into modification of contacts (e.g. contact tracing and isolation). The figure scenarios (school openness levels) seems more interpretable as X% scaling of relevant age groups' contacts so I'll stick to those for now.