gemlib.mcmc.discrete_time_state_transition_model.right_censored_events_mh

gemlib.mcmc.discrete_time_state_transition_model.right_censored_events_mh#

gemlib.mcmc.discrete_time_state_transition_model.right_censored_events_mh(incidence_matrix, transition_index, t_range, count_max=1, name=None)#

Update right-censored events for DiscreteTimeStateTransitionModel

In a partially-complete epidemic, we may wish to explore the space of events that _might_ have occurred, but are not apparent because of some detection event that is yet to occur. This MCMC kernel performs a single-site “add/delete” move, as described (in continuous time) in O’Neill and Roberts (1999).

Parameters:

  • incidence_matrix (ndarray[tuple[int, ...], dtype[_ScalarType_co]]) – the state-transition graph incidence matrix

  • transition_index (int) – the index of the transition in incidence_matrix to update

  • t_range (tuple[int, int]) – the time-range for which to update censored transition events. Typically this would be [s, num_steps) where s < num_steps for num_steps the total number of timesteps in the model.

  • count_max (int) – the max number of transitions to add or delete in any one update

  • name (str | None) – name of the kernel

Returns:

An instance of SamplingAlgorithm with members init(tlp_fn: LogProbFnType, position: Position, initial_conditions: Tensor) and step(tlp_fn: LogProbFnType, cks: ChainAndKernelState, seed: SeedType, initial_conditions: Tensor). initial_conditions is an extra keyword argument required to be passed to the init and step functions.

Return type:

SamplingAlgorithm

References

  1. Philip D O’Neill and Gareth O Roberts (1999) Baysian inference for partially observed stochastic epidemics. _Journal of the Royal Statistical Society: Series A (Statistics in Society), 162 :121–129.

  2. Jewell _et al._ (2023) Bayesian inference for high-dimensional discrete-time epidemic models: spatial dynamics of the UK COVID-19 outbreak. Pre-print arXiv:2306.07987

Contents