Construct Likelihood Function for Statistical Models

Source: R/construct_likelihood.R

Build likelihood, log-likelihood, and deviance functions from scratch without external dependencies. Supports profile likelihood and likelihood-based inference.

construct_likelihood(model = NULL, data = NULL, formula = NULL,
  family = gaussian(), method = c("auto", "numeric", "analytic"))

Arguments

model

Fitted model object (lm, glm, etc.) or NULL to build from scratch

data

Data frame (required if model is NULL)

formula

Model formula (required if model is NULL)

family

Family object for GLMs (default: gaussian)

method

Method for likelihood construction: "auto", "numeric", "analytic"

Value

List containing likelihood functions and methods

Details

Based on:

  • Pawitan (2001). In All Likelihood. Oxford University Press.

  • Venzon & Moolgavkar (1988). A method for computing profile-likelihood-based confidence intervals. Applied Statistics, 37(1), 87-94.

  • Murphy & van der Vaart (2000). On profile likelihood. JASA, 95(450), 449-465.