Skip to contents

SIRD-GA log-likelihood for daily reported cases and/or deaths

Source: R/sird_ga_incidence.R
sird_ga_loglik_inc.Rd

Kalman-filter approximate log-likelihood of the SIRD-GA state-space model when the observations are daily increments (new reported cases, new reported deaths), as recorded by routine surveillance systems. Daily increments are linear in the augmented state \(Z_k=(X_k, X_{k-1})\), so the linear-Gaussian structure is preserved. Transition quantities are computed in compiled code (RK4 on the joint system for the ODE solution, the resolvent and the Lyapunov equation of the state-noise covariance).

Usage

sird_ga_loglik_inc(
  theta,
  cases = NULL,
  deaths = NULL,
  N,
  use = c("both", "cases", "deaths"),
  Delta = 1,
  nsub = 10L,
  full = FALSE,
  lambda_t = NULL
)

Arguments

theta

Named vector with lambda, gamma, alpha, s0, i0, r0, p, tau2, q, tau2_d.

cases, deaths

Daily counts for intervals 1, ..., n (NA allowed). Either may be NULL when use excludes it.

N

Population size.

use

Which streams enter the likelihood: "both", "cases" or "deaths".

Delta

Observation interval.

nsub

RK4 steps per interval.

full

If TRUE, return the full list (fitted one-step-ahead means and standard deviations, ODE path) instead of the scalar.

lambda_t

Optional vector of transmission rates, one per observation interval; when given it overrides theta[["lambda"]] (piecewise-constant transmission).