inference: MCMC based inference of the parameter values given the...

In MJomaba/flu-evidence-synthesis: Flu Evidence Synthesis: Analysing Effectiveness of Vaccine Strategies

MCMC based inference of the parameter values given the different data sets

Description

MCMC based inference of the parameter values given the different data sets

Usage

inference(demography, ili, mon_pop, n_pos, n_samples, vaccine_calendar,
  polymod_data, initial, parameter_map, age_groups, age_group_map,
  risk_group_map, risk_ratios, lprior, lpeak_prior, nburn = 0,
  nbatch = 1000, blen = 1, depth = 3, abs_err)

Arguments

demography	A vector with the population size by each age 0,1,..
ili	The number of Influenza-like illness cases per week
mon_pop	The number of people monitored for ili
n_pos	The number of positive samples for the given strain (per week)
n_samples	The total number of samples tested
vaccine_calendar	A vaccine calendar valid for that year
polymod_data	Contact data for different age groups
initial	Vector with starting parameter values or one can pass the results of a previous fit to continue from those results
parameter_map	Optional mapping parameter (by description and age group) to the relevant index in the initial vector. Needed parameters are: epsilon (ascertainment) with a separate value per data age group, transmissibility, psi (infection from outside sources), susceptibility (with a value per age group) and log of initial_infected population parameter_mapping.
age_groups	Optional age groups upper limits used in your model and data. If you use different age groups for the model and the data you need to provide a age_group_map instead.
age_group_map	Optional age group mapping from model age groups to data age groups (age_group_mapping)
risk_group_map	Optional risk group mapping from model risk groups to data risk groups (risk_group_mapping). This parameter is not needed if only one risk group is modelled
risk_ratios	A matrix with the fraction in the risk groups. The leftover fraction is assumed to be low risk. (stratify_by_risk)
lprior	Optional function returning the log prior probability of the parameters. If no function is passed then a flat prior is used.
lpeak_prior	Optional function to include prior knowledge on the peak time and height. This function should accept a time and height (no. of infected in the population) and return a log likelihood value for those values.
nburn	Number of iterations of burn in
nbatch	Number of batches to run (number of samples to return)
blen	Length of each batch
depth	Obsolete, use abs_err instead. The depth of the likelihood approximation. Higher values result in better approximations. The default value is 3.
abs_err	The absolute error of the likelihood approximation. Lower values are more precise at the cost of performance. Default value is 1e-5.

Details

The method we use here combines data from numerous sources that are then used to compute the likelihood of the predicted number of influenza cases in a given week. Given the data and the likelihood function we use MCMC to obtain the posterior distribution of the parameters of an underlying epidemiological model (see also: infectionODEs).

When running inference there are four main steps needed are 1) prepare the data, 2) load a vaccination calendar (as_vaccination_calendar) 3) decide on parameterisation of the model (https://blackedder.github.io/flu-evidence-synthesis/modelling.html) and 4) run the inference using this function.

The initial parameters vector should contain values for the parameters (in order):

• Ascertainment probabilty for each age group (epsilon)

• Outside infection (psi)

• Transmissibility

• Susceptibility for each age group

• Initial number of infections (log transformed)

If your model is more complex and the number of age groups and risk groups are different between the epidemiological model (vaccination calendar) and the influenza data then you need to provide (one or more of) the following extra variables to the function: parameter_map (see also: parameter_mapping), age_group_map (see also: age_group_mapping) and risk_group_map (see also: risk_group_mapping). See https://blackedder.github.io/flu-evidence-synthesis/inference.html for more details.

Value

Returns a list with the accepted samples and the corresponding llikelihood values and a matrix (contact.ids) containing the ids (row number) of the contacts data used to build the contact matrix.

See Also

infectionODEs; age_group_mapping; risk_group_mapping; parameter_mapping; https://blackedder.github.io/flu-evidence-synthesis/inference.html

MJomaba/flu-evidence-synthesis documentation built on April 26, 2022, 11:12 p.m.