Estimates the variance-covariance matrix of log-incidence estimates
by modelling residual autocorrelation with an exponential decay model.
Usage
vcov_from_residuals(
data,
mu,
sigma,
max_lag = 10,
min_alpha = 0.01,
max_alpha = 2
)
Arguments
- data
Numeric vector: observed case counts I_t
- mu
Numeric vector: estimated log-incidence (log lambda_t)
- sigma
Numeric vector: estimated standard error of log lambda_t
- max_lag
Integer: maximum lag for ACF estimation (default: 10)
- min_alpha
Positive number: minimum decay rate (default: 0.01)
- max_alpha
Positive number: maximum decay rate (default: 2.0)
Value
List with:
vcov_matrix: T x T estimated VCOV matrix for log lambda_t
alpha: fitted decay parameter
acf_obs: observed ACF of Pearson residuals
acf_fit: fitted ACF values
residuals: Pearson residuals used
times: time indices
Details
This function was generated by Qwen3-235B-A22B-2507
Examples
T <- 50
mu <- 5 + 0.05 * (1:T) + stats::arima.sim(list(ar = 0.8), n = T, sd = 0.3)
sigma <- rep(0.2, T)
lambda <- exp(mu)
data <- stats::rpois(T, lambda)
result <- vcov_from_residuals(data, mu, sigma, max_lag = 8)
dim(result$vcov_matrix) # Should be T x T
#> [1] 50 50
