McMasterPandemic: Pandemic Model

Understanding `make_beta` Under Two-Dose Vaccination Model

Icats are the names of the infected states -- Ia, Ip, Im, Is
Icat_prop_vec is Ca, Cp, (1 - iso_m) * Cm, (1 - iso_s) * Cs)
beta_0 is beta0 * 1/N * Icat_prop_vec
vax_cat is a list of state variable suffixes that identify vaccination status unvax, vaxdose1, vaxprotect1, vaxdose2, vaxprotect2 for a two-dose model -- every state variable is categorized in this way
vax_trans_red is a 5-by-5 matrix (corresponding to the vax_cat status')
set the vax_cat == vaxprotect1 row to 1 - vax_efficacy_dose1
set the vax_cat == vaxdose2 row to 1 - vax_efficacy_dose1
set the vax_cat == vaxprotect2 row to 1 - vax_efficacy_dose2
all other rows at 1 so far
beta_0 gets updated to kron(vax_trans_red, t(beta_0))
beta is a matrix with 5 rows (same as beta_0) and n_state columns -- fill the corresponding beta_0 values in beta, leaving the remaining columns all zero (so we can leave these out in a more efficient implementation)

testcats are ??

Understanding `add_updated_vaxrate` Under Two-Dose Model

Compute the vax_rate list with one scalar element per dose
Set the dimensions of the problem
epi_states -- "S" "E" "Ia" "Ip" "Im" "Is" "H" "H2" "ICUs" "ICUd" "D" "R" "X" "V"
asymp_cat -- "S" "E" "Ia" "Ip" "R"
vax_cat -- "unvax" "vaxdose1" "vaxprotect1" "vaxdose2" "vaxprotect2"
Initialize two identical block matrices (vax_block_dose1 and vax_block_dose2) with all zeros, with rows and columns each corresponding to one of the epi_states
Fill in vax_rate$dose_n into each of the following elements of vax_block_dose_n (where _n is either 1 or 2)
Diagonal elements corresponding to each asymp_cat
Any element from any asymp_cat to "V" (if "V" is present)
Place each block into ratemat:
vax_block_dose1 -- unvax categories to vaxdose_1 categories
vax_block_dose2 -- vaxprotect1 categories to vaxdose_2 categories

Understanding `make_vaxrate` Under Two-Dose Model

vax_rate is a vector of rates, each of which will be used to fill many rate matrix elements
asymp_unvax_regex -- "^(S|E|Ia|Ip|R)_.*unvax"
asymp_vaxprotect1_regex -- "^(S|E|Ia|Ip|R)_.*vaxprotect1"
dose1 = vax_prop_first_dose * vax_doses_per_day / asymp_unvax_N
dose2 = (1 - vax_prop_second_dose) * vax_doses_per_day / asymp_unvax_N

(aside) Understanding `condense_state` for `return_type == "named_vector"`

Important: This is not the same as condense.pansim, which is where the bottleneck could be -- so the tidyverse stuff here is unlikely to be the cause of my problem. On the other hand, this stuff could be a big contributor to the vax slowness and I bet this could be sped up simply with R refactoring (no C++ necessary)

Pass a state vector or subset
Seems like we are looking for states that start with these characters:
states -- c("S", "E", "I", "H", "ICU", "R", "D")
states_regex -- "^S" "^E" "^I" "^H" "^ICU" "^R" "^D"
Convert input to a one-row data frame (but looks like it can have more rows in some cases -- oh ... I see ... probably these cases are when we have a time-series of simulation values for each state variable) and then go tidyverse on it
Select using states_regex
Add an obs_number column -- which probably means 'time-step' in the case of condensing simulations
Pivot longer so that obs_number and var refer to each value
Split up the variable name into state and subcat by splitting on _
Make subcat a factor (why?)
Group by obs_number and subcat and summarise, which has the effect of one row per subcat (e.g. vaxprotect1), and sum up the values
Pivot wider so that there is one row per obs_number and a column for each subcat
Finally get rid of obs_number so that the only columns represent subcat
Return the requested return type
Answer
asymp_unvax_N = sum('S_unvax', 'E_unvax', 'Ia_unvax', 'Ip_unvax', 'R_unvax')
asymp_vaxprotect1_N = sum('S_vaxprotect1', 'E_vaxprotect1', 'Ia_vaxprotect1', 'Ip_vaxprotect1', 'R_vaxprotect1')