Models, Data and Inference for Socio-Technical Systems
Connect probability models and statistics for real systems: queueing, regression, experiments, plus a weekly contest on misused statistics in the media.
Good system decisions depend on both probabilistic models and data, and this doctoral class links the two. It opens with applied probability, using puzzles like the 3-door and broken stick problems, then moves through queueing into statistics, including regression and design of experiments. A weekly contest rewards the worst misuse of statistics in the media, which keeps things lively.
What You Will Achieve
- Develop axiomatic models using applied probability
- Analyze queues using Markov processes
- Test hypotheses and fit regression models
- Design experiments and assess threats to validity
Program Overview
Applied Probability
The first half builds models from assumptions.
- 3-door problem
- Random incidence
- Spatial models
- Markov processes
Derived Distributions and Statistics
A bridge links the two ways of modeling.
- Derived distributions to statistics
- Psychology of queueing
- Weibull distribution
Statistics and Experiments
The second half tests models against data.
- Hypothesis testing
- Multiple regression
- Bayesian and frequentist statistics
- Design of experiments
Course Completion
Job Outlook
These skills fit research on engineering systems, operations analysis, policy evaluation and data science.
Good for: Doctoral students and quantitative analysts with prior probability who model systems under uncertainty.
Keep in mind: From spring 2007, with no videos. Software is your own choice, though later assignments need substantial computation. The two suggested books are not required, and the project and contest depended on the class.
Bottom line: Lively and well balanced, and the contest winners are an entertaining lesson in reading statistics critically.
Frequently Asked Questions
What is the weekly contest?
Students nominate the worst use or abuse of statistics in the media, and some winning entries, with their source articles, are published.
Which software should I use?
Your own choice. The syllabus considers MATLAB with the Statistics Toolbox a good option, and it warns that coding in C or FORTRAN would be inefficient.
What are the two types of models?
Axiomatic models built from stated assumptions, without statistics, and statistical models that rely on data, hypothesis tests and parameter estimates.



There are no reviews yet.