Create BinaryVariable state object

[giovannic]

Models often include binary variables that are not mutually exclusive (like State objects). These are currently modelled with Variables using 1 and 0. Investigate and integrate performant alternatives.

[giovannic]

Categorical variable?

[slwu89]

If I wanted to create a model of, for example, house to house transmission where each agent's current house belongs to a discrete set where all possible combinations of infection states and houses are possible "states" for the agent, I should wait for this update?

[giovannic]

Currently it's only possible using a few Variable objects as indicators for each state and doing something like:

contains_susceptible = individual::Variable$new(house, rep(0, population))
contains_infected = individual::Variable$new(house, rep(0, population))
contains_recovered = individual::Variable$new(house, rep(0, population))

process <- function(api) {
  # do something with houses which have susceptible and infected
  both <- (api$get_variable(house, contains_susceptible) == 1) & (api$get_variable(house, contains_infected) == 1)
  ...
}

This is not ideal. It sounds like this issue would suit you better but I can't give an estimate on when that will be available.

[slwu89]

Is this something I could help out with? It looks like this is a special case of issue #47 that you mentioned, maybe something like Categorical could cover all these discrete non-mutually exclusive aspects of state?

[giovannic]

Ok great!

So we intend Categorical to be mutually exclusive, and MultiCategory to be non-mutually exclusive.

Categorical is going to be introduced in #46, currently WIP on feat/variable_refactor. It's not fully tested and documented yet though :/

If you're keen to use MultiCategory now, you could branch off of feat/variable_refactor and implement MultiCategory. We could then it in once the refactor is complete. I have added some extra TODO notes on #47 if you want to go down that route.

[slwu89]

Oh cool, that's very clear! Thanks for pointing me to the extra notes on #47, I'll need to study the refactored API a bit before starting those.