search

mrc-ide / hypatia

Individual-based COVID transmission model
https://mrc-ide.github.io/hypatia
Other
0 stars 0 forks source link

Model varying Rt #18

Open giovannic opened 4 years ago

giovannic commented 4 years ago

We should allow a time series of Rt to be specified through tt_R0 and R0 parameters.

This involves integrating interpolation between the time points for R0.

Tests:

giovannic commented 4 years ago

Do interpolations beforehand. Then in the process do a lookup

OJWatson commented 3 years ago

Rough outline/thoughts for this. Users will want to provide values of R0 in one of two ways:

  1. Single value (current approach)
  2. Multiple values with a value of R0 for each timestep.
  3. (They may want to provide an R0 value at specific timesteps and want us to interpolate them, but the end result is that we will have a value of R0 for each timestep).

If we pass to squire::parameters_explicit_SEEIR a vector of the timesteps ( seq(0, end_timestep, 1) ) and a vector of R0 values then we will get back in pars a vector of beta values for each timestep. These need to then be correctly used for each timestep in this line:

https://github.com/mrc-ide/hypatia/blob/main/R/infection_process.R#L30

lambda <- pars$beta * rowSums(m %*% diag(inf_ages))

We need to change that so that instead of using pars$beta we use pars$beta[current_timestep]. We should be able to get current_timestep from the individual api (api$get_timestep()).

The simplest way I can see then is to make it so that if the user provides one value of R0, we adapt pars$beta so that it has the same length as the number of timesteps. And then add in some checks to make sure that if the user provides fewer than the number of timesteps worth of betas we throw an error to tell them to provide and R0 value for every timestep.

(Later on we can add some interpolation functions to hypatia to catch examples where the user provides say tt_R0 = c(0,10,20,30) and R0 = c(4,3,2,1) and we do some linear interpolation to make it so that they have passed as.numeric(mapply(rep, 4:1, diff(c(0,10,20,30,end_timestep)))) for the R0 etc. But for the time being let's assume they either provide 1 value or a vector)