Skip to contents

Estimate individual-level load variation

Source: R/model_shedding.R
load_variation_estimate.Rd

This option accounts for variation in the total shedding load per case by modeling individual shedding loads as Gamma distributed with mean equal to the average load_per_case and a coefficient of variation to be estimated.

Usage

load_variation_estimate(
  cv_prior_mu = 1,
  cv_prior_sigma = 0,
  modeldata = modeldata_init()
)

Arguments

cv_prior_mu

Mean of the truncated normal prior for the coefficient of individual-level variation. Default is a CV of 1, which means that approximately 60% of individuals shed less than the mean, and approximately 10% of individuals shed less than 10% of the mean.

cv_prior_sigma

Standard deviation of the truncated normal prior for the coefficient of individual-level variation. If this is set to zero, the coefficient of variation is fixed to the mean.

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

Note that measurement noise and individual-level load variation might not be jointly identifiable. This is why the coefficient of variation of the individual-level load is fixed by default.

Also note that the accuracy of the variation model depends on the estimated number of cases to be on the right scale (which in turn depends on the assumed load_per_case to be roughly correct). This is because in the variation model, the population-level coefficient of variation (CV) of shedding loads is proportional to the individual-level CV divided by the square root of the number of cases. If for example the assumed load_per_case is too small, the number of cases will be overestimated, which has effects in two directions:

  • the individual-level CV will be overestimated (especially if the prior on the individual-level CV is weak)

  • the population-level CV will be underestimated (especially if the prior on the individual-level CV is strong)

See also

Other load variation models: load_variation_none()