Infer the prevalence of disease from incidence estimates and population size.

Log-scaled incidence estimates are used to generate samples of incidence. These are convolved with the infectivity profile (as a set of discrete distributions) and the result is evaluated as a fraction of population. The result is fitted to a logit-normal (using moments) and quantiles produced from samples.

Usage

infer_prevalence(
  modelled = i_incidence_model,
  pop = i_population_data,
  ip = i_discrete_ip,
  bootstraps = 1000,
  seed = Sys.time(),
  ...
)

Arguments

modelled

Model output from processing the raw dataframe with something like poission_locfit_model

A dataframe containing the following columns:

time (ggoutbreak::time_period + group_unique) - A (usually complete) set of singular observations per unit time as a `time_period`
incidence.fit (double) - an estimate of the incidence rate on a log scale
incidence.se.fit (positive_double) - the standard error of the incidence rate estimate on a log scale
incidence.0.025 (positive_double) - lower confidence limit of the incidence rate (true scale)
incidence.0.5 (positive_double) - median estimate of the incidence rate (true scale)
incidence.0.975 (positive_double) - upper confidence limit of the incidence rate (true scale)

Any grouping allowed.

pop

The population data must be grouped in the same way as modelled.

A dataframe containing the following columns:

population (positive_integer) - Size of population

Any grouping allowed.

ip

A discrete distribution representing the probability of detection of disease on a given day after infection. This will not necessarily sum to one, but each entry will not exceed the asymptomatic fraction.

A dataframe containing the following columns:

boot (anything + default(1)) - a bootstrap identifier
probability (proportion) - the probability of new event during this period.
tau (integer + complete) - the days since the index event.

Must be grouped by: boot (exactly).

A default value is defined.

bootstraps

the number of samples to take at each time point. This will be rounded up to a whole multiple of the infectivity profile distribution length.

seed

a random number seed for reproducibility

...

not used

Value

the modelled input with additional proportion columns

Examples

tmp = ggoutbreak::england_covid %>%
  dplyr::filter(date < "2021-01-01") %>%
  time_aggregate(count=sum(count)) %>%
  ggoutbreak::poisson_locfit_model(window=21) %>%
  ggoutbreak::infer_prevalence(
     pop = 56489700,
     # N.B. timing is difficult in this example, as we are using cases.
     ip = ggoutbreak::covid_test_sensitivity
  )

if(interactive()) {
  plot_prevalence(tmp)
}