New model - Stochastic SEIR with population structure

[roberthinch]

A new stochastic SEIR model with population structure. The model is discrete time with daily stepping. The individuals in the compartments are assumed to be well-mixed, thus every individual experiences the same force of infection. Infection events are also assumed to be independent, thus the number of new infections at each time step is binomially distributed. For individuals, other transitions are assumed to be geometrically distributed, thus the number of people moving between compartments is also binomial distributed. Population and interventions are modelled using the same classes as model_default(). Currently the implementation does not support time-dependent parameters (except via interventions) or vectorised parameters. The default number of stochastic replicates is 1,000, and the runtime for the first example of model_default() (i.e. age-structured UK population with school closures run for 600 days) takes a few seconds (so about 100 times slower than the deterministic model).

Manual documentation and tests have been adapted from the existing ones for model_default().

Issue #260

[roberthinch]

Note, there is one un-related test failure in test-population which fails for me locally in the production code.