Advanced statistical models

Silvia Cagnone; Luca Trapin; Alessandra Luati; Aldo Gardini; Silvia De Nicolò;

  • Date:

    10 JANUARY
    -
    07 MAY 2025
     from 10:00 to 17:00
  • Event location: Aula IV, Department of Statistical Sciences, Via Belle Arti 41 (STAT PhD Classes Virtual Room on need) - In presence and online event

  • Type: Cycle 40 - Mandatory Courses

Aims: To introduce advanced modelling strategies for independent data or data characterized by spatial or temporal dependence including model selection, inference, prediction.
Learning outcomes: At the end of the course the students will have an appreciation of the different modelling strategies for data characterized by various kinds of dependencies and will be able to deal with estimation issues, to perform model selection and to apply the models to real data situations.

Course content

Module 1. Generalized linear (GLM) and latent variable models (GLLVM) (Prof. S. Cagnone)

- GLM:

  • Introduction to GLM
  • Model Specification
  • Inference
  • Extentions and Applications

- GLLVM:

  • Introduction to GLLVM
  • Model specification for continuous and categorical data
  • Inference
  • Applications

Module 2. Models for panel data (Prof. L. Trapin)
- Large N, Small T:

  • Static panels: pooled, fixed effect, random effect estimators
  • Dynamic panels: endogeneity, IV, and gmm estimators

- Large N, Large T:

  • Interactive fixed effect estimators
  • Grouped panels

- Nonlinear panels

Module 3. Advanced time series analysis: Time-varying-parameter models (Prof. A. Luati)

  • Nonlinear non-Gaussian models
  • Maximum likelihood estimation
  • Asymptotic theory

Module 4. Specification and Inference for models with cross-sectional (spatial) dependence (Prof. Aldo Gardini; Prof. Silvia De Nicolò)

  • Fundamental concepts of Bayesian hierarchical modeling with an introduction to Latent Gaussian Models
  • Incorporating correlation structures into models using Gaussian Markov Random Fields (GMRF) and intrinsic GMRF
  • Focus on spatio-temporal models: Random Walk processes, Intrinsic Conditional Auto-Regressive (ICAR) priors, and their interactions
  • Introduction to Small Area Estimation models (S. De Nicolò).