Skip to contents

Estimate infection noise

Source: R/model_infections.R
infection_noise_estimate.Rd

This option estimates noise in the infection process, i.e. implements a stochastic renewal model. This allows for variation in the number of new infections generated at each time step, which can often speed up model fitting.

Usage

infection_noise_estimate(
  overdispersion = TRUE,
  overdispersion_prior_mu = 0.1,
  overdispersion_prior_sigma = 0,
  modeldata = modeldata_init()
)

Arguments

overdispersion

If TRUE (default), new infections are modeled as Negative Binomial distributed. If FALSE, new infections are modeled as Poisson distributed.

overdispersion_prior_mu

Prior (mean) on the overdispersion parameter of the Negative Binomial. The default of 0.1 corresponds to 10% overdispersion. It is also the limit of the coefficient of variation (CV) of infections as the infection incidence becomes large.

overdispersion_prior_sigma

Prior (standard deviation) on the overdispersion parameter of the Negative Binomial. If this is set to zero (default), the overdispersion will be fixed to the prior mean and not estimated.

modeldata

A modeldata object to which the above model specifications should be added. Default is an empty model given by modeldata_init(). Can also be an already partly specified model returned by other EpiSewer modeling functions.

Value

A modeldata object containing data and specifications of the model to be fitted. Can be passed on to other EpiSewer modeling functions to add further data and model specifications.

The modeldata object also includes information about parameter initialization (init), meta data (.metainfo), and checks to be performed before model fitting (.checks).

Details

The level of overdispersion is often unidentifiable from a single time series of measurements. This is why the overdispersion is fixed by default.

For complicated reasons, MCMC sampling of high infection numbers is faster with a certain degree of overdispersion. Thus, if modeling large waves, assuming some overdispersion can also make sense for computational reasons. The effects on the estimated transmission dynamics are often minimal.

The priors of this component have the following functional form:

  • overdispersion parameter of the Negative Binomial: Truncated normal

See also

Other infection noise models: infection_noise_none()