B1967
Title: Predictive performance test based on the exhaustive nested cross-validation for high-dimensional data
Authors: Zhaoxia Yu - University of California Irvine (United States)
Hernando Ombao - King Abdullah University of Science and Technology (KAUST) (Saudi Arabia)
Iris Ivy Gauran - King Abdullah University of Science and Technology (Saudi Arabia) [presenting]
Abstract: Cross-validation is a fundamental algorithmic technique with widespread applications, including the estimation of prediction error, regularization parameter tuning, and selecting competing predictive models, among others. However, its behavior can be intricate, influenced by a myriad of complex factors. A novel approach is introduced based on exhaustive nested cross-validation, designed for straightforward application with minimal assumptions about the underlying data distribution. Moreover, our proposed method can generate valid confidence intervals for determining the difference in prediction error between two model-fitting algorithms. We address concerns regarding computational complexity by devising a highly efficient expression for the cross-validation estimator. Our study also delves into strategies to enhance statistical power within high-dimensional scenarios while controlling the Type I error rate. To illustrate the practical utility of our method, we apply it to an RNA sequencing study and demonstrate its effectiveness in the context of biological data analysis.