COVID wastewater model to match macpan1.5 implementation

canmod / macpan2

Rebuilding https://github.com/mac-theobio/McMasterPandemic/

https://canmod.github.io/macpan2/

GNU General Public License v3.0

2 stars 0 forks source link

COVID wastewater model to match macpan1.5 implementation #156

Open stevencarlislewalker opened 8 months ago

stevencarlislewalker commented 8 months ago

macpan2 implementation here

[x] initial expression list and naive defaults
[x] hazard correction
[x] gamma-kernel convolution for reports
[ ] variant takeover
[x] time-variation of beta0

On the likelihood branch, update the model library specification and calibration script to use real data and reflect the changes here.

[x] data prep for observed time series data (reports and W)
[ ] can/should mobility data be included in updated calibration
[x] relevant variable transformations
[x] use new convolution functionality to compute reports (mp_tmb_insert_reports)
[x] use explicit likelihood functionality (mp_neg_bin, mp_normal)
[x] test hazard correction functions when all infectious states are zero
[ ] can/should phenomenological heterogeneity be included

stevencarlislewalker commented 8 months ago

Related to #94

stevencarlislewalker commented 8 months ago

info from macpan1.5 is here (note this is not available publicly)

stevencarlislewalker commented 8 months ago

simplest possible hazard correction example here.

jfree-man commented 1 month ago

Hazard correction works conditional on having at least one infectious individual at time zero, otherwise S.E flow is NaN.

stevencarlislewalker commented 1 month ago

Hazard correction works conditional on having at least one infectious individual at time zero, otherwise S.E flow is NaN.

Thanks @jfree-man. I think this is a minimal example.

si = (mp_tmb_library("starter_models", "si", package = "macpan2")
  |> mp_tmb_update(default = list(I = 0))
)
(si
  |> mp_hazard()
  |> mp_simulator(5L, "I")
  |> mp_trajectory()
)

I also think that this is fixed on the likelihood branch. Could be fixed elsewhere, but the key to the fix I think is the proportions function on the cpp side.

jfree-man commented 1 month ago

Default parameter values, simulated data from macpan1.5, and observed data for calibration are available on the likelihood branch here. In the future, data should be accessible from the package and shouldn't necessarily require running a data fetching script every time it's needed.