LIKpoisson_glmm: Likelihood of the Poisson GLMM

In johannabertl/ApproxML: Approximate maximum likelihood estimation

Description

The likelihood of the parameters beta and sigma in a Poisson GLMM is computed. The integral is calculated with Gauss-Hermite quadrature.

Usage

LIKpoisson_glmm(beta, sigma, x, y)

LOGLIKpoisson_glmm(beta, sigma, x, y)

Arguments

beta	scalar
sigma	scalar
x	vector, regressor
y	vector, regressand

Details

The function dpois is used to evaluate the conditional likelihood, as it is fast (implemented in C) and can handle large values of y, that can otherwise be problematic in the evaluation of factorial(y).

The loglikelihood can sometimes be obtained even when the likelihood evaluates to zero (numerically).

Functions

• LOGLIKpoisson_glmm: LOGLIKpoisson_glmm directly sums the log of the likelihood terms that are computed in the same way as by the function LIKpoisson_glmm.

Examples

x = 1:3
y = c(2,4,6)
beta = 1
sigma = 1

betavec = seq(0, 5, by =0.1)
likvec = numeric(length(betavec))
for(i in 1:length(likvec)) likvec[i] = LIKpoisson_glmm(betavec[i], sigma, x, y)
plot(betavec, likvec)

johannabertl/ApproxML documentation built on May 22, 2019, 2:19 p.m.