EM algorithm for INAD model estimation

Fits INAD models using the Expectation-Maximization algorithm. This is an alternative to direct likelihood optimization.

Usage

em_inad(
  y,
  order = 1,
  thinning = "binom",
  innovation = "pois",
  blocks = NULL,
  max_iter = 200,
  tol = 1e-07,
  alpha_init = NULL,
  theta_init = NULL,
  tau_init = NULL,
  nb_inno_size = NULL,
  safeguard = TRUE,
  verbose = FALSE
)

Arguments

y

Integer matrix with n_subjects rows and n_time columns.

order

Model order (1 or 2). Order 0 does not require EM.

thinning

Thinning operator: "binom", "pois", or "nbinom".

innovation

Innovation distribution: "pois", "bell", or "nbinom".

blocks

Optional integer vector of length n_subjects for block effects.

max_iter

Maximum number of EM iterations.

tol

Convergence tolerance for log-likelihood change.

alpha_init

Optional initial values for alpha parameters.

theta_init

Optional initial values for theta parameters.

tau_init

Optional initial values for tau parameters.

nb_inno_size

Size parameter for negative binomial innovation (if used).

safeguard

Logical; if TRUE, use step-halving when likelihood decreases.

verbose

Logical; if TRUE, print iteration progress.

Value

A list with class "inad_fit" containing estimated parameters.

Details

For Gaussian and CAT EM entry points, see em_gau and em_cat. For CAT specifically, fit_cat() supports na_action = "em" for orders 0/1 and na_action = "marginalize" for order 2 missing-data fits.

See also

em_gau, em_cat, fit_inad, fit_cat

Examples

set.seed(1)
y <- simulate_inad(
  n_subjects = 50,
  n_time = 5,
  order = 1,
  thinning = "binom",
  innovation = "pois",
  alpha = 0.25,
  theta = 2
)
fit <- em_inad(y, order = 1, thinning = "binom", innovation = "pois", max_iter = 20, tol = 1e-6)
fit$log_l
#> [1] -439.3326