B0826
Title: Significance tests based on neural networks with applications to genetic association studies
Authors: Xiaoxi Shen - Texas State University (United States) [presenting]
Chang Jiang - University of Florida (United States)
Lyudmila Sakhanenko - Michigan State University (United States)
Qing Lu - Michigan State University (United States)
Abstract: Despite the great success of applications of neural networks in many different fields, such as natural language processing and image recognition, there is a lack of research that focuses on the interpretation of neural network models. Two hypothesis testing methods based on neural networks with one hidden layer will be introduced to conduct significance tests of input features. The asymptotic distributions for both test statistics are simple, so it is easy to apply in real data analysis. The validity of the asymptotic distributions is investigated via simulations, and we applied our proposed tests to perform a genetic association analysis on the sequencing data from Alzheimer's Disease Neuroimaging Initiative (ADNI).
