pipette
ENEnglish

Robust Global Testing for Mean Functions in Sparse and Dense Functional and Longitudinal Data

Behdad Mostafaiy, Mustafa Abduljabbar Dawood

Preprint

En palabras de los autores

We develop a robust quadratic score test for a specified mean function in functional and longitudinal data. The construction combines a convex loss with subject-level scores while retaining dependence among repeated measurements. An explicit score-centering condition ensures that the target remains the population mean. We establish the null distribution and Sobolev separation rates under sparse and dense sampling, identifying a transition from a nonparametric testing rate to the root-n boundary. A Gaussian multiplier bootstrap remains valid across these regimes, including settings with non-Gaussian quadratic limits. For implementation, we use a finite spline approximation and estimate the curvature through symmetric score increments. Under the stated conditions, the feasible procedure preserves null calibration and attains a consistency separation rate matching the lower bound over model classes containing a Gaussian submodel. Simulations illustrate the performance of squared, Huber, and absolute losses under regular errors and contamination, and an analysis of longitudinal CD4 counts demonstrates the method.

Resultado principalEl resumen no menciona limitaciones.

Apareció: viernes, 30 de octubre. arXiv. Preprint, todavía sin revisión por pares.