Skip to contents

As befits a model-fitting function, the package defines a nearly complete set of methods for "nestedLogit" objects:

  • print() and summary() print the results for each of the submodels.
  • update() re-fits the model, allowing changes to the model formula, data, subset, and contrasts arguments.
  • coef() returns the coefficients for the predictors in each dichotomy.
  • vcov() returns the variance-covariance matrix of the predictors.
  • predict() computes predicted probabilities for the response categories, either for the cases in the data or for arbitrary combinations of the predictors; the latter is useful for producing plots to aid interpretation.
  • glance() and tidy() are extensions of broom::glance.glm() and broom::tidy.glm() to obtain compact summaries of a "nestedLogit" model object.
  • plot() provides basic plots of the predicted probabilities over a range of values of the predictor variables.
  • models() is an extractor function for the binary logit models in the "nestedLogit" object

These are supplemented by various methods for testing hypotheses about and comparing nested logit models:

  • anova() provides analysis-of-deviance Type I (sequential) tests for each dichotomy and for the combined model. When given a sequence of model objects, anova() tests the models against one another in the order specified.
  • Anova() uses car::Anova() to provide analysis-of-deviance Type II or III (partial) tests for each dichotomy and for the combined model.
  • linearHypothesis() computes Wald tests for hypotheses about coefficients or their linear combinations.
  • logLik() returns the log-likelihood and degrees of freedom for the nested-dichotomies logit model.
  • Through logLik(), the AIC() and BIC() functions compute the Akaike and Bayesian information criteria model-comparison statistics.