On the Semiparametric Efficiency of a Class of Functional Response Models for Between-Subject Attributes

• Activity Number: 55 - Statistical methods for data from single cell technologies
• Type: Contributed
• Date/Time: Sunday, August 8, 2021 : 3:30 PM to 5:20 PM
• Sponsor: Royal Statistical Society
• Author(s): Jinyuan Liu* and Tian Chen and Tuo Lin and Xinlian Zhang and Xin Tu
• Companies: UCSD and The University of Toledo and University of California, San Diego and UCSD and University of California, San Diego
• Keywords: Asymptotic Efficiency Bound; Between-subject Attribute; Hilbert Space; High-throughput Data; Semiparametric Model; U-statistics based Generalized Estimating Equations (UGEE)

Abstract

Motivated by difficulties to correctly specify model assumptions confronting huge amounts of unstructured data, semiparametric models are receiving growing interest, where asymptotic efficiency bounds are of fundamental importance to quantify efficiency loss resulting from fewer model restrictions. We discuss efficiency bounds for a class of semiparametric functional response models (FRM), which overcomes the limitations within the predominant within-subject regression paradigm, by providing a timely tool for the intrinsic between-subject attributes summarizing high-dimensional data. Despite its growing applications, efficiencies of parameter estimators have not been carefully studied. Leveraging the efficiency theory for within-subject models, we develop both concepts and results for asymptotic efficiencies in this broader regression setting for modeling between-subject attributes. As between-subject outcomes have become increasingly popular in the burgeoning fields of biomedical research as effective summary metrics for high-throughput data, the established concepts and results will propel new and wider applications of the FRM and related semiparametric models.