M. Bugallo Porto, D. Morales, N. Salvati, F. Schirripa Spagnolo
M-quantile (MQ) regression presents a promising alternative to mixed models for small area estimation (SAE), though several formal aspects still require further exploration. Firstly, we prove the consistency of the area-specific MQ coefficients, which in MQ models serve as the equivalent of random effects in mixed models. Secondly, we investigate the optimal selection of the robustness parameter for bias correction, making a theoretical contribution to the field and improving outlier detection. Moreover, we address several issues related to inference and diagnosis in MQ models, proposing bootstrap techniques to approximate the distribution of area-specific MQ coefficients and robustness parameters. As for the mean squared error estimation, a general method accounting for total variability is proposed, based on an approximation of the residuals' distribution in MQ models. Key directions for future research are outlined.
Keywords: Small Area Estimation, M-quantile model, Robust Inference, Bias correction, Bootstrap
Scheduled
Ramiro Melendreras Awards
June 10, 2025 7:00 PM
Sala VIP Jaume Morera i Galícia