lrtest: Likelihood-ratio test for nested regression models

In R4VN: Health Data Analysis and Publication-Ready Reporting

Likelihood-ratio test for nested regression models

Description

Compares two or more nested likelihood-based regression models. Objects returned by R4VN logistic() and poisson() can be supplied directly.

Usage

lrtest(..., digits = 3, p_digits = 3, show = TRUE, console = FALSE)

Arguments

...	Two or more nested fitted models, ordered from the smaller model to progressively larger models. R4VN statistical results containing a fitted model in raw$model are accepted directly.
digits	Number of decimal places for likelihood and LR statistics.
p_digits	Number of decimal places for p-values.
show	Logical; open the formatted result in the Viewer. Default TRUE.
console	Logical; also print the result in the Console. Default FALSE.

Details

When more than two models are supplied, comparisons are sequential: M1 versus M2, then M2 versus M3, and so on.

The models must use the same outcome, analytic observations, weights, offsets/exposure definition, and likelihood family/link, and each larger model must contain the smaller model.

Supported fits include ordinary likelihood-based glm models such as logistic and Poisson regression, MASS::glm.nb() negative-binomial models, survival::coxph() Cox models, and survival::survreg() parametric survival models.

Quasi-likelihood models are not supported. R4VN models fitted with vce = "robust" or vce = "cluster" are also rejected because the classical likelihood-ratio chi-square test is model-likelihood inference, not robust covariance inference.

For ordinary linear regression use the nested-model F test rather than lrtest().

Value

An object of class r4vn_stat. The unformatted comparison table is stored in result$raw$table; the backward-compatible result$raw$comparison table is also retained.

See Also

logistic(), poisson()

Examples

set.seed(2026)
d <- data.frame(
  y = factor(rbinom(250, 1, .35), levels = 0:1,
             labels = c("No", "Yes")),
  age = rnorm(250, 45, 12),
  sex = factor(sample(c("Female", "Male"), 250, TRUE)),
  treatment = factor(sample(c("No", "Yes"), 250, TRUE))
)

m1 <- logistic(y, c.age, i.sex, i.treatment,
               data = d, event = "Yes", show = FALSE)
m2 <- logistic(y, c.age, i.sex*i.treatment,
               data = d, event = "Yes", show = FALSE)

lrtest(m1, m2, show = FALSE)

R4VN documentation built on Sept. 30, 2026, 5:13 p.m.