Different networks for different demographics

Custom networks significantly increase the flexibility of metawards. You can use them to create models for different countries. And you can use them to use the concept of metapopulations to model different environments.

For example, we can use a custom network to model home, school and work, and then use different demographics to model students, teachers and everyone else. This way you could explore how closing schools could impact disease spread.

This concept is explored in detail in the quick start guide. The key takeaway you should have now, having worked through this tutorial as well as the quick start, is that you can create;

• different Disease objects that contain different disease stages and parameters,

• different Demographic objects that represent different groups of individuals, each of which can experience different Disease objects (and thus advance along different pathways,

• who can each experience metapopulation movements on different Network networks (built flexibly via the Ward / Wards objects),

• where each ward in each network can have its own parameters, thereby enabling modelling of ward-local behaviour for different demographics at different disease stages in different networks.

This flexibility for the model is then enhanced by building custom iterators that enable you to write your own code to advance individuals from one disease stage to the next, and control how susceptible individuals are infected.

You can build custom mixers to control how the force of infections calculated for different demographics are mixed and merged together, thereby controlling how they interact (from not interacting, to evenly interacting, via custom interactions for the interaction matrix).

You can build custom movers to control vertical movements of individuals between demographics, that complement the horizontal movement of individuals along disease stages.

And you can write custom extractors to analyse the data from the simulation on the fly, and write whatever data you wish out to disk or database.

And, from here you can creata custom user parameters that control all of this model, and write scan/design files that can run multiple simulations to scan through these custom parameters, as well as (nearly) all in-built and disease parameters, optionally repeating runs to check for statistical variance.

And (finally!) you can do all of this from within a single Python or R script using the Python or R APIs, or you can run from the command line, or from within a batch script that will automatically parallelise the runs over a supercomputing cluster.