Syllabus summary
Course identity
Mathematical Statistics II (MATH 75163) — 3 credits, three contact hours a week of graduate lecture and workshop.
Purpose
Treat a statistical procedure as a decision rule with assumptions and operating characteristics rather than a mechanical route to a p-value: derive how a procedure is built, calibrate what it actually does, stress it where its assumptions fail, and compare exact, asymptotic, resampling, robust, and Bayesian answers to the same question.
Outcomes overview
- Formulate statistical inference as a decision problem with an estimand, action space, loss function, procedure, and risk.
- Define and calculate size, level, power, type I error, type II error, and power functions, and explain what each quantity does and does not establish.
- Apply the Neyman-Pearson lemma and monotone likelihood-ratio arguments to construct most powerful and uniformly most powerful tests.
- Use the duality between tests and confidence sets to derive and interpret exact and approximate procedures.
- Derive and compare likelihood-ratio, score, and Wald tests, including their regularity assumptions and finite-sample limitations.
- Handle nuisance parameters using conditioning, profiling, invariance, or asymptotic approximation when appropriate.
- Derive central results for normal linear models using projections and quadratic forms, including t, F, regression, and analysis-of-variance procedures.
- Explain and evaluate family-wise error control, false-discovery-rate control, multiple comparisons, and introductory selective-inference concerns.
- Construct and justify randomization, permutation, and exact conditional procedures from the assignment or sampling mechanism.
- Use bootstrap and jackknife methods, and assess their calibration through theory and simulation.
- Apply foundational nonparametric rank procedures and compare their assumptions, targets, and efficiency with parametric alternatives.
- Evaluate robustness using influence, contamination, breakdown, M-estimation, and sandwich-style variance ideas.
- Derive and compare Bayes rules, minimax rules, admissible procedures, shrinkage estimators, and predictive decisions in foundational settings.
- Distinguish evidence against a null model, posterior model probability, predictive performance, and scientific importance.
- Use R to estimate operating characteristics, compare procedures, diagnose approximation failure, and reproduce inferential results.
- Audit AI-generated mathematical and computational work by checking assumptions, theorem conditions, edge cases, code execution, and independent derivations.
- Communicate a defensible inferential conclusion with explicit assumptions, calibration claims, sensitivity results, and limitations.
Materials and access
All course materials are provided on this site at no cost. Open sources in support:
- MIT OpenCourseWare 18.655 Mathematical Statistics, Spring 2016 (Peter Kempthorne, MIT OpenCourseWare) — open graduate reference. availability and licence not yet confirmed.
- Penn State STAT 415 Introduction to Mathematical Statistics (Department of Statistics, Penn State Eberly College of Science) — open inference reference. availability and licence not yet confirmed.
- The R Project for Statistical Computing (The R Foundation for Statistical Computing) — computing environment. availability and licence not yet confirmed.
- Quarto (Posit PBC and the Quarto contributors) — reproducible authoring tool. availability and licence not yet confirmed.
Logistics
The public materials for this course are in development. Dates, policies, assessments, and readings on this site are provisional and may change, and the term shown is a planning assumption rather than a scheduled section. The learning management system remains authoritative for section logistics and graded details.