Yiming Shi

Program: Biomedical Informatics and Data Science

Current advisor: Lei Liu, PhD

Undergraduate university: Fudan University, 2019

Enrollment year: 2022

Research summary
Model integrating robust covariance estimators with Poisson regression and application in microbiome study

Microbiome research often encounters challenges due to the complex nature of microbiome data. One issue is that the abundance counts in microbiome data exhibit overdispersion or heteroscedasticity, which can lead to incorrect inferences if not properly addressed. To tackle this, we consider several approaches, including bootstrap covariance estimation and sandwich robust covariance matrix estimators. Our research demonstrates that integrating robust covariance estimators with Poisson regression yields accurate model estimations for both coefficients and covariance. This approach offers comparable power to many complex covariance estimation frameworks while simplifying the model and reducing the number of variables to be estimated, thus enhancing the robustness of model fitting. Our findings are supported by an extensive series of simulation studies, highlighting the efficacy of robust covariance estimators in addressing the unique challenges of microbiome data analysis.

Graduate publications
Xu J, Farsad HL, Hou Y, Barclay K, Lopez BA, Yamada S, Saliu IO, Shi Y, Knight WC, Bateman RJ, Benzinger TLS, Yi JJ, Li Q, Wang T, Perlmutter JS, Morris JC, Zhao G. 2023 Human striatal glia differentially contribute to AD- and PD-specific neurodegeneration. Nat Aging, 3(3):346-65. PMCID: PMC10046522

Shi Y, Li H, Wang C, Chen J, Jiang H, Shih YT, Zhang H, Song Y, Feng Y, Liu. 2023 A flexible quasi-likelihood model for microbiome abundance count data. Stat Med, 42(25):4632-43. PMCID: PMC11045296