Yunnan Xu

Research goal is to explore effects of hurricane Sandy on healthcare utilization of elder adults. • Investigate the depressing effect of hurricane sandy on elder adults’ health condition further influencing the healthcare utilization using survey data and medicare data with data manipulation and statistical analyses performed in SAS

Consulting work at Statistical Applications & Innovations Group (SAIG) of Virginia Tech • Helped non-statisticians with statistical concepts, statistical software (R, JMP, SAS) and plotting data, and collaboration on design of experiments, analysis, and interpretations of analysis results. • Held weekly walk-in consulting hours

Recitation Leader • Gave review sessions, supervised students on recitation assignments, and graded assignment

A Hybrid Approach for Prediction of Event Times in Double-Blind Clinical Trials A hybrid of greedy algorithm and permutation is employed to allocate subjects, followed by construction of the survival function for event time prediction, which consists of change point detection, parameter estimation using maximum likelihood, and extension of parametric tail. Work included algorithm development and refinement, data simulation, program implementation, performance testing and comparison using R, presentation of work and results, and writing a manuscript. Manuscipt: A Nonparametric Approach for Prediction of Event Times in Double-blind Clinical Trials

Exploration of correlations among biological pathways, diseases and genes • Extract and aggregate data from ToxCast database and National Toxicology Program (NTP) database in SQL, fit regression models to assay data and animal experiment data, and apply correlation analysis to find correlations between diseases and biological pathways and discover correlated biological pathways and corresponding correlated genes

The lab instructor role of teaching Chemistry Lab courses at Department of Chemistry required giving lectures, supervising students performing experiments, grading assignments and exams, advising students during office hours, and developing exam questions.

Department of Statistics Topic: Statistical Methods for In-session Hemodialysis Monitoring • Develop a peak-preserving baseline correction algorithm for Raman spectra using iterative smoothing-spline root error adjustment, and apply it to Raman spectra from hemodialysis patients • Develop a two-sample test on mixed data, i.e., data with functional variables (e.g., the entire Raman spectrum) and scalar variables (e.g., intensities at peak locations of interesting biomarkers on the Raman spectrum), with p-values of the test statistics computed using asymptotic distributions • Sparse logistic regression with a LASSO-type penalty on functional data with identification of regions of difference Virginia Tech Transportation Institute Topic: Event identification within naturalistic driving data • Clean and manipulate naturalistic driving data, and identify crash and non-crash outcomes though logistic regression on principal components of acceleration records in a certain time period