IEDA5270
Engineering Statistics
2019–present
Course Snapshot
- Institution
- HKUST, Department of Industrial Engineering and Decision Analytics
- Period
- 2019–present
- Audience
- Graduate students in engineering and data analytics, and advanced undergraduates with strong quantitative backgrounds.
- Prerequisites
- Probability theory, linear algebra, calculus, and basic programming/statistical computing.
This course develops a rigorous statistical toolkit for engineering data analytics, covering probability foundations, random-sample theory, estimation and inference, regression modeling, regularization, generalized linear models, and modern classification methods.
Learning Outcomes
- Formulate statistical inference problems and interpret assumptions in engineering contexts.
- Construct and evaluate point estimators, hypothesis tests, and confidence sets.
- Build and diagnose linear and generalized linear regression models.
- Apply model selection, regularization, and cross-validation for predictive performance.
- Compare and deploy classification methods based on data characteristics and evaluation metrics.
Lecture Modules
Materials
Review of Probability Theory
Events, conditional probability, random variables, expectation, transformations, and multivariate distributions as the foundation for inference.
Open HandoutProperties of a Random Sample
Sampling distributions, moment generating functions, order statistics, LLN/CLT, and delta-method approximations.
Open HandoutPrinciples of Data Reduction
Sufficient, minimal sufficient, and complete statistics, with factorization and exponential-family perspectives.
Open HandoutPoint Estimation
Method of moments, maximum likelihood, Fisher information, Cramer-Rao bounds, and unbiased estimation under sufficiency/completeness.
Open HandoutHypothesis Testing
Test construction, simple and composite hypotheses, UMP tests, likelihood-ratio methods, and sequential testing principles.
Open HandoutConfidence Set
Confidence-set construction via pivotal quantities and test inversion, plus asymptotic, bootstrap, and Bayesian intervals.
Open HandoutRegression Models
Multiple linear regression, least squares theory, inference, nested model tests, diagnostics, Box-Cox transforms, spline, and robust methods.
Open HandoutModel Selection and Regularization
Subset selection, ridge and lasso shrinkage, model criteria (Adjusted R2/Cp/AIC/BIC), and cross-validation workflows.
Open HandoutGeneralized Linear Model
Logistic and Poisson regression, exponential-family formulation, estimation algorithms, and GLM inference/testing.
Open HandoutClassification Methods
LDA/QDA, k-NN, tree-based ensembles, and SVM methods for supervised classification tasks.
Open Handout