Courses Taught by Xu Shi

BIOSTAT681: Introduction to Causal Inference

• Graduate level
• Residential
• Fall term(s) for residential students;
• 3 credit hour(s) for residential students;
• Instructor(s): Xu Shi (Residential);
• Prerequisites: None
• Advisory Prerequisites: Biostats 601, 602, 650, and 651
• Description: This course is designed to introduce students to basic of causal inference, including potential outcomes, counterfactuals, confounding, mediation, and instrumental variables. We will explore the identification and estimation of causal effects via the use of principle stratification, marginal structural models, and directed acyclic graphs.
• Learning Objectives: At the end of the course, students should be able to -Define concepts including potential outcomes, confounding, and mediation. -Explain the purpose of randomization for causal inference. -Explain the concept of principal stratification/casual association and use the relevant statistical methods to make causal inference under the casual association paradigm. -Explain the concepts of direct and indirect effects/casual effects and use the relevant statistical methods to make causal inference under the casual effects paradigm. -Understand how instrumental variables can be related to other causal inference approaches. -Consider the use of directed acyclic graphs (DAGs) to define causal concepts and derive conditions of identifiability.
• Syllabus for BIOSTAT681

BIOSTAT881: Topics In Advanced Causal Inference

• Graduate level
• Residential
• Winter term(s) for residential students;
• 3 credit hour(s) for residential students;
• Instructor(s): Xu Shi (Residential);
• Prerequisites: BIOSTAT 601, BIOSTAT 602, BIOSTAT 650, BIOSTAT 651, BIOSTAT 653, BIOSTAT 801, and BIOSTAT 802 (concurrent also accepted).
• Description: This course covers statistical theory and methodology for drawing causal conclusions from observational and experimental data. We will cover theoretical foundations including DAGs and SEMs, followed by special topics, which may include instrumental variable analysis, causal inference in high dimensions, and causal inference with longitudinal data.
• Learning Objectives: At the end of the course the students will be able to: 1. Translate a scientific question into a causal contrast to be estimated. 2. Derive graphical models for investigating the conditions under which the causal contrasts of interest are identified from data collected under specific study designs. 3. Formulate adequate structural models for making inference about the causal contrasts of interest. 4. Implement simulations appropriate for investigating the properties of causal estimators.
• Syllabus for BIOSTAT881

EPID731: Analysis Of Electronic Health Record (ehr) Data

• Graduate level
• Residential
• Summer term(s) for residential students;
• 1 credit hour(s) for residential students;
• Instructor(s): Staff, Xu Shi, (Residential);
• Offered Annually during Summer term
• Last offered Summer 2025
• Prerequisites: None
• Advisory Prerequisites: Quantitative training, familiarity with traditional regression methods, basic epidemiologic principles, and working knowledge of R. The course will be instructed with minimal mathematics formulas and will include comprehensive examples to facilitate a bro
• Undergraduates are allowed to enroll in this course.
• Description: To gain knowledge of the process of cleaning and abstracting EHR data to create analytic datasets, attain a broader understanding the secondary use of EHR data for research, with a focus on epidemiologic principles including the role of study design, bias, and generalizability
• Learning Objectives: This short course will offer an overview of modern analytical methods and research applications using EHR data, with a specific focus on epidemiologic inferences. Upon completion of the course, participants will i) gain knowledge of the process of cleaning and abstracting EHR data to create analytic datasets, ii) attain a broader understanding of the opportunities and challenges of the secondary use of EHR data for research, with a focus on epidemiologic principles including the role of study design, bias, and generalizability, iii) explore and gain hands-on experience using EHRs from Michigan Medicine, and iv) be prepared to generate and further explore new questions and perspectives.