Park, Jun Young2022-08-292022-08-292020-05https://hdl.handle.net/11299/241295University of Minnesota Ph.D. dissertation. May 2020. Major: Biostatistics. Advisor: Mark Fiecas. 1 computer file (PDF); x, 108 pages.Studies using imaging data of the human brain, henceforth called neuroimaging, have various types of study designs, and provide a number of data types including, but not limited to, behavioral data, functional/structural magnetic resonance imaging (fMRI/sMRI), electroencephalogram (EEG), positron emission tomography (PET), and even genetics data. Neuroimaging data is often high dimensional and yields several layers of complex correlation structures that statistical modeling and inference need to account for in order to improve statistical power. Popular statistical methods include reducing the data to summary statistics, conducting a massive univariate analysis on the summary statistics, and dimension reduction of the summary statistics using principal component analysis (PCA) or independent component analysis (ICA). Despite their simplicity, these approaches (i) often ignore important features of the brain not captured by those summary statistics, (ii) are less powerful by not accounting for correlation structures, and (iii) miss common characteristics present in multiple data types. In this thesis, I develop statistical modeling and inference procedures to alleviate the issues and apply the methods to different types of neuroimaging and genomics data to show their performances and possible extensions.enThesis or Dissertation