Fit Transmission Model Using PMCMC

Runs a particle MCMC fit of the PREVAIL transmission model using pre-generated parameters and a serosurvey dataset. This function performs a PMCMC run using the dust2 engine for likelihood evaluation and the monty framework for inference. The function does not run simulation post-fitting — instead, it returns everything needed for follow-up analysis or second-stage fitting.

Usage

fit_transmission_model(
  parameters,
  serodata,
  prior,
  stochastic = F,
  fitted_parameters = c("reporting_rate", "R0_modifier"),
  initial = NULL,
  domain = matrix(c(0, 2, 0, 2), nrow = 2, byrow = T),
  vcv = diag(0.1, length(fitted_parameters)),
  n_steps = 1000,
  n_chains = 1,
  n_particles = 1
)

Arguments

parameters

A named list of model parameters, as produced by custom_data_process_wrapper.

serodata

A data.frame with columns time and serosurvey, where serosurvey is a list-column containing numeric vectors of observed seropositivity by age.

prior

A prior distribution object created using monty_dsl.

stochastic

TRUE or FALSE. Determines whether to run the transmission_model() as stochastic or deterministic. By default uses FALSE to increase speed.

fitted_parameters

Character vector of parameter names to fit (e.g., c("reporting_rate", "R0_modifier")).

initial

A named list of initial parameter values, or NULL to default all to 1.

domain

A matrix with two rows (lower and upper) and columns named for each fitted parameter. Defines parameter bounds.

vcv

A variance-covariance matrix for the random walk proposal. Should be square with dimensions equal to number of parameters.

n_steps

Integer. Number of MCMC steps to perform (default = 1000).

n_chains

Integer. Number of independent MCMC chains (default = 1).

n_particles

Integer. Number of particles used in the particle filter (default = 1).

Value

A named list containing:

samples

The output from monty_sample, including sampled parameters and log-likelihoods.

posterior

The final posterior object used for inference.

summarised_draws

Summary statistics (e.g., mean, median, quantiles) for each parameter from the posterior samples.

all_states

The full simulated state trajectories from the model for both base and fitted parameter sets.

plots

A list of ggplot2 objects including trace plots, log-likelihood plot, combined trace-loglik plot, and seropositivity plot: trace, loglik, combined, sero.

Examples

if (FALSE) { # \dontrun{
params <- PREVAIL::custom_data_process_wrapper(iso = "KEN", disease = "measles", R0 = 15)
serodata <- data.frame(time = 2500, serosurvey = I(list(rep(0.85, 17))))
prior <- monty::monty_dsl({
  reporting_rate ~ Normal(mean = 1, sd = 5)
  R0_modifier ~ Normal(mean = 1, sd = 1)
})
fit <- fit_transmission_model(parameters = params, serodata = serodata, prior = prior)
fit$plots$combined
} # }