Examinando por Autor "Elena, Castilla"

Mostrando 1 - 2 de 2

A New Robust Approach for Multinomial Logistic Regression With Complex Design Model
(IEEE, 2022-06-29) Elena, Castilla; Pedro J., Chocano
Robust estimators and Wald-type tests are developed for the multinomial logistic regression based on ϕ -divergence measures. We compute the influence function of the proposed estimators and tests and discuss some consequences. Their robustness is illustrated by an extensive simulation study and two real examples.
On the choice of the optimal tuning parameter in robust one-shot device testing analysis
(Springer, 2023) Elena, Castilla; Pedro J., Chocano
During the last decade, considerable work has been carried out in one-shot device analysis and, in particular, in robust methods based on divergences, which improve the classical inference based on the maximum likelihood estimator (MLE) or likelihood ratio test. The estimators and tests developed by this approach depend on a tuning parameter . The choice of is, however, one of the main drawbacks of this perspective. In this paper, given a data set, we study different methods for the choice of the “optimal” tuning parameter including the iterative-Warwick and Jones (IWJ) algorithm (Basak et al. [8]) or the minimization of some loss functions of the observed data. While IWJ algorithm seems to be a good approach for low and moderate contamination, some simulations do suggest that minimizing the mean absolute error of the observed probabilities is as least as efficient as the IWJ algorithm for high contamination, avoiding heavier computations.