Compare nested models fitted to equivalent interdep data

Performs a likelihood-ratio test for two nested glmmTMB models fitted to separate interdep_data objects. Unlike anova.glmmTMB(), the model calls do not need to refer to the same R object. The function instead checks that the prepared data contain the same original observations before comparing the models.

Usage

compare_interdep_models(full, restricted, alpha = 0.05)

Arguments

full

The full (larger) fitted glmmTMB model.

restricted

The restricted (smaller) fitted glmmTMB model.

alpha

Significance level used for the printed interpretation.

Value

An anova-style data frame containing model degrees of freedom, information criteria, log-likelihoods, the likelihood-ratio statistic, and its chi-squared p-value. Printing the result adds a cautious interpretation based on alpha.

Details

Both model calls must use named interdep_data objects. The function checks the original, non-.i_ columns, structural dyad metadata, outcome values and missingness, fitted row identities, model family and link, maximum-likelihood estimation, and model convergence.

These checks establish that the models use equivalent observations. They cannot establish that one model is mathematically nested within the other. The caller remains responsible for supplying a genuinely restricted model and its corresponding full model. The usual chi-squared reference distribution may also be inappropriate when tested variance parameters are on the boundary.

Examples

if (requireNamespace("glmmTMB", quietly = TRUE)) {
  restricted_data <- prepare_interdep_data(
    example_dyadic_crosssectional,
    group = coupleID,
    member = personID,
    role = gender
  )
  full_data <- restricted_data

  restricted_model <- glmmTMB::glmmTMB(
    satisfaction ~ 1 + us(1 | coupleID),
    data = restricted_data
  )
  full_model <- glmmTMB::glmmTMB(
    satisfaction ~ gender + us(1 | coupleID),
    data = full_data
  )

  compare_interdep_models(
    full = full_model,
    restricted = restricted_model
  )
}
#> Likelihood-ratio test for nested models fitted to equivalent interdep data
#> Mathematical nesting is assumed and cannot be verified from the data alone.
#> 
#>                  Df    AIC    BIC  logLik deviance  Chisq Chi Df Pr(>Chisq)   
#> restricted_model  3 917.29 926.98 -455.64   911.29                            
#> full_model        4 908.52 921.45 -450.26   900.52 10.765      1   0.001034 **
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#> 
#> Under the assumed nesting and chi-squared reference distribution, the test provides evidence that `restricted_model` fits the data worse than `full_model` (likelihood-ratio test: χ²(1) = 10.77, p = 0.001).