Testing distance-to-centroid dependent covariate effects in directional quantile bivariate regression: glycemic data analysis
S. Rodríguez Pastoriza, J. Roca Pardiñas, Ó. Lado Baleato
Demographic covariates like age or sex do more than shift biomarker location as they can reshape the distribution’s geometry and joint dependence. Ignoring this reconfiguration across the quantile spectrum risks producing biased reference regions. We propose a dual inferential framework to test if a covariate justifies adjusting a bivariate region. Our approach combines a test based on the Wild Bootstrap with a quantile specific Bayesian Information Criterion. These tools are complementary because while the Bootstrap offers formal guarantees for the distribution's body, the BIC maintains power at extreme quantiles where resampling degrades. Applied to diabetes research, age reconfigures the relationship between fasting plasma glucose and glycated hemoglobin. This allows robust estimation of individualized regions, helping identify metabolic discordance in patients at risk whose profiles are atypical despite univariate values within standard ranges.
Keywords: Covariate effect testing, quantile-specific inference, wild bootstrap, BIC, glycemic biomarkers, diabetes
Scheduled
Biostatistics I
September 4, 2026 11:10 AM
Aula 24
Other papers in the same session
I. Arostegui Madariaga, I. Barrio Beraza, A. Iparragirre Letamendia
M. C. Iglesias Pérez, T. Rodríguez Rodríguez
M. Martínez Pizarro, M. J. Rufo Bazaga, J. D. Hernández Guillén, A. Moreno Díaz
A. Ruiz Vega, M. Á. Montero Alonso, J. D. Luna del Castillo
A. Fanjul Hevia, J. C. Pardo Fernández