2017/11/20 by Ronak Mehta, Hyunwoo J. Kim, Mehta, Ronak +9
Biochemistry, Genetics and Molecular Biology · Mathematics · Medicine · #Bioinformatics and Genomic Networks #Data-Driven Disease Surveillance #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.1711.07575
openalex publication_date 2017/11/20 · openalex created_date 2022/09/05 · openalex updated_date 2026/07/28
Recent results in coupled or temporal graphical models offer schemes for\nestimating the relationship structure between features when the data come from\nrelated (but distinct) longitudinal sources. A novel application of these ideas\nis for analyzing group-level differences, i.e., in identifying if trends of\nestimated objects (e.g., covariance or precision matrices) are different across\ndisparate conditions (e.g., gender or disease). Often, poor effect sizes make\ndetecting the differential signal over the full set of features difficult: for\nexample, dependencies between only a subset of features may manifest\ndifferently across groups. In this work, we first give a parametric model for\nestimating trends in the space of SPD matrices as a function of one or more\ncovariates. We then generalize scan statistics to graph structures, to search\nover distinct subsets of features (graph partitions) whose temporal dependency\nstructure may show statistically significant group-wise differences. We\ntheoretically analyze the Family Wise Error Rate (FWER) and bounds on Type 1\nand Type 2 error. On a cohort of individuals with risk factors for Alzheimer's\ndisease (but otherwise cognitively healthy), we find scientifically interesting\ngroup differences where the default analysis, i.e., models estimated on the\nfull graph, do not survive reasonable significance thresholds.\n