Advanced Multivariate Analysis

A.Y. 2025/2026
6
Max ECTS
40
Overall hours
SSD
SECS-S/05
Language
English
Learning objectives
Undefined
Expected learning outcomes
Undefined
Single course

This course can be attended as a single course.

Course syllabus and organization

Single session

Responsible
Lesson period
Third trimester
Course syllabus
The course begins with a review of core concepts in quantitative analysis, including Monte Carlo simulation, the logic of inference, and the assumptions and diagnostics of linear regression models. It then extends the linear model to include categorical predictors, interactions, and conditional hypotheses, before introducing Generalized Linear Models such as logistic regression for binary outcomes, Poisson and Negative Binomial models for count data, and models for proportions. Attention is then given to model comparison and predictive validation. The second part of the course focuses on advanced approaches, covering multilevel and hierarchical models (random intercepts, random slopes, and cross-classified structures), longitudinal and panel data analysis, and applications in political and social science. The final part of the course introduces Bayesian methods, including Bayesian multilevel modeling. The course concludes with frontier topics such as text-as-data methods, regularization and penalized regression (ridge, LASSO, elastic net), nonparametric and semi-parametric methods (kernel regression and generalized additive models), and an bridge to machine learning approaches relevant for social science applications.
Prerequisites for admission
A prior introductory course in statistics and some basic experience with R (or another programming language) are recommended. Students should be comfortable with descriptive statistics, probability, and simple regression models. While no advanced mathematical background is required, the course assumes a willingness to engage actively with quantitative methods and computational tools.
Teaching methods
The course adopts a hands-on, student-centered approach that combines lectures with interactive activities, coding exercises, and group discussions. Statistical concepts are introduced through real-world applications rather than abstract proofs, with extensive use of the R programming language for data analysis. Assigned readings are studied collaboratively on the Perusall platform to foster peer-to-peer learning, while GitHub is used to support reproducibility and teamwork in coding exercises. The teaching methods emphasize participation, inclusivity, and self-learning, preparing students to apply advanced multivariate techniques to original research questions in social and political science.
Teaching Resources
Core teaching materials for the course will be provided directly by the instructor in the form of lecture notes and slides. A useful reference text is Regression and Other Stories by Gelman, Hill, and Vehtari (Cambridge University Press, 2020), which complements the course by covering many of the foundational and applied methods. Additional readings, datasets, and resources will be circulated during the course to support the advanced topics.
Assessment methods and Criteria
Student performance will be assessed through a combination of continuous and final evaluation. Continuous evaluation is based on active participation in reading assignments (via the Perusall platform) and completion of group exercises with reproducible code shared on GitHub. The final evaluation is based on an individual capstone project, which may consist of either a replication study of a published article or a draft of an original research paper. The final grade reflects both the quality of individual work and active engagement with peers. Innovative self-learning tools (including AI-assisted coding support) may also be used to complement the assessment.
SECS-S/05 - SOCIAL STATISTICS - University credits: 6
Lessons: 40 hours
Professor: De Angelis Andrea
Professor(s)