Speaker
Description
Identifying predictors associated with specific response categories in multinomial logistic regression is a challenging task. It is furthermore complex in a high-dimensional setting, where the number of covariates is higher than the number of units. To address the variable selection in high dimensional domain and in the presence of multinomial models with unordered responses, we propose a two-step ranking-based approach for category-specific variable selection. The method relies on marginal multinomial regressions, in which each covariate is separately regressed on the response variable, thereby enabling the identification of predictors relevant to individual categories.
The comparison of the proposed approach with standard penalized techniques demonstrates the parsimony of the final model (after variable selection) and the accuracy of its predictive performance. A further advantage of the proposed approach is the computational efficiency and scalability, which make the method well-suited for models involving a large number of predictors and response categories.
| Special/ Invited session | itENBIS (sessione invitata della Società Italiana di Statistica) - Organizer: Prof. Amalia Vanacore |
|---|---|
| Classification | Both methodology and application |
| Keywords | Multinomial regression model, high dimension, variable selection |