SDGLM (version 0.4.0)

dglm_likelihood: Calculate Log-Likelihood for DGLM

Description

Element-wise log-likelihood for Dynamic Generalized Linear Models with linear predictor eta = X * beta.

Usage

dglm_likelihood(
  y,
  X,
  beta,
  family = c("poisson", "pareto", "gamma"),
  alpha_gamma = 2,
  lambda_min = 1e-06
)

Value

Numeric vector of length n; -Inf for illegal observations.

Arguments

y

Numeric response vector (length n).

X

Numeric design matrix (n x p).

beta

Numeric coefficient vector (length p).

family

Character distribution choice: "poisson", "pareto", or "gamma".

alpha_gamma

Shape parameter for Gamma (default 2).

lambda_min

Lower bound for Pareto shape lambda (default 1e-6).

Examples

set.seed(123)
X <- matrix(rnorm(100 * 3), 100, 3)
beta <- c(0.5, -0.2, 0.1)

# Poisson
y_pois <- rpois(100, lambda = exp(X %*% beta))
ll_pois <- dglm_likelihood(y_pois, X, beta, family = "poisson")

# Gamma
y_gamma <- rgamma(100, shape = 3, rate = exp(X %*% beta))
ll_gamma <- dglm_likelihood(y_gamma, X, beta, family = "gamma", alpha_gamma = 3)