Research Interest
High-dimensional data analysis, time series analysis, imaging statistics
Conferences and Presentations
Song, H., Eloyan, A., & Lee, Y. "Leverage of Functional Connectivity and Effective Connectivity by Selective Inference with Sample Splitting and fMRI Data." ENAR 2024 Spring Meeting, Baltimore, MD, March 2024 (Upcoming), accepted.
Causal Inference Approaches for Effective Connectivity with fMRI Data
Research Assistant | Prof. Youjin Lee and Prof. Ani Eloyan | Dec. 2022 - Present | Brown University
Pre-processed fMRI data with 88 * 88 * 64 * 976 dimension from the ADNI dataset, including inhomogeneity correction, co-registration and registration to the template, spatial smoothing by R, and mapped the data to 120 time courses representing functional activity from regions of interest
Calculated Pearson's correlation coefficients with Fisher's Z transformation to identify strongly correlated regions as functional connectivity analysis, and tested the correlations and visualized the p-values through 120 * 120 heatmap
Fitted vector autoregressive model and inferred conditional Granger causality between two time courses, conditioning on strongly correlated time courses
Leveraged the functional connectivity and effective connectivity by sample splitting to address the challenges of choosing potential confounding factors and the high-dimensional problem in fMRI data
Preliminary results, both in theory and simulations, showed the proposed method is asymptotically valid under certain conditions, effectively controlling type-I error rates
Two Sample Test for High Dimensional Data
Research Program | Prof. Jin-Ting Zhang | May 2021 - Jul. 2021 | NUS
Attended weekly seminars covering high-dimensional data analysis topics
Reviewed literature and delivered a forty-five-minute presentation on "Effects of High Dimension: Two-Sample Problems"
Comparison of VaR Approaches in China A-Share Market during COVID-19
Univ. of Cambridge and SWUFE Joint Research Program | Jan. 2021 - May 2021 | Dr. Matthias Dörrzapf
Performed exploratory data analysis to financial market data before and after COVID-19 widespread; analyzed and visualized volatility by standard deviations
Computed the Value-at-Risk by GARCH, EGARCH, Simulations; calculated Expected Shortfall values; assessed their accuracy based on significance levels of back-testing results