API Documentation

The likelihood.model package provides a modular framework for constructing likelihood functions from independent contributions. This page summarizes the core API.

Core Class: `likelihood_contr_model`

The central abstraction is the likelihood_contr_model S3 class, which represents a likelihood model built from independent contributions.

Constructor

likelihood_contr_model(
  obs_type,           # Type of observation: "exact", "right_censored", etc.
  distribution,       # Distribution object (e.g., weibull_distribution())
  data = NULL,        # Optional data frame
  ...
)

Generic Functions

The package implements several generic functions that any likelihood model must support:

Function	Description
`loglik(model, theta, data)`	Compute log-likelihood at parameter values
`score(model, theta, data)`	Compute score function (gradient)
`fisher_info(model, theta, data)`	Compute Fisher information matrix
`obs_loglik(model, theta, data)`	Per-observation log-likelihood contributions

Observation Types

Different observation types contribute differently to the likelihood:

Exact Observations

For fully observed data, the contribution is the log-density:

\ell_i(\theta) = \log f(x_i; \theta)

Right-Censored Observations

For right-censored data (e.g., survival analysis), the contribution uses the survival function:

\ell_i(\theta) = \delta_i \log f(t_i; \theta) + (1 - \delta_i) \log S(t_i; \theta)

where $\delta_i$ is the event indicator.

Interval-Censored Observations

For interval-censored data, the contribution is:

\ell_i(\theta) = \log[F(b_i; \theta) - F(a_i; \theta)]

Distribution Interface

Distribution objects must implement:

# Density function
pdf(dist, x, theta)

# Cumulative distribution function
cdf(dist, x, theta)

# Survival function
sf(dist, x, theta)

# Hazard function (optional)
hf(dist, x, theta)

Built-in Distributions

weibull_distribution() - Two-parameter Weibull
exponential_distribution() - Exponential (special case of Weibull)
lognormal_distribution() - Log-normal distribution

Integration with algebraic.mle

The package is designed to work seamlessly with algebraic.mle:

library(likelihood.model)
library(algebraic.mle)

# Create likelihood model
model <- likelihood_contr_model(
  obs_type = "right_censored",
  distribution = weibull_distribution()
)

# Fit using MLE
fit <- mle(model, data = my_data, start = c(shape = 1, scale = 1))

# Access results
coef(fit)           # Parameter estimates
vcov(fit)           # Variance-covariance matrix
confint(fit)        # Confidence intervals

Full Reference

For complete API documentation, including all function signatures and examples, see the pkgdown documentation.