Training and Research

PhD school courses/classes - 2023/2024

Please note: Additional information will be added during the year. Currently missing information is labelled as “TBD” (i.e. To Be Determined).

1. PhD students must obtain a specified number of CFUs each year by attending teaching activities offered by the PhD School.
First and second year students must obtain 8 CFUs. Teaching activities ex DM 226/2021 provide 5 CFUs; free choice activities provide 3 CFUs.
Third year students must obtain 4 CFUs. Teaching activities ex DM 226/2021 provide 2 CFUs; free choice activities provide 2 CFUs.
More information regarding CFUs is found in the Handbook for PhD Students: https://www.univr.it/phd-vademecum

2. Registering for the courses is not required unless explicitly indicated; please consult the course information to verify whether registration is required or not. When registration is actually required, no confirmation e-mail will be sent after signing up. Please do not enquiry: if you entered the requested information, then registration was silently successful.

3. When Zoom links are not explicitly indicated, courses are delivered in presence only.

4. All information we have is published here. Please do not enquiry for missing information or Zoom links: as soon as we get new information, we will promptly publish it on this page.

Teaching Activities ex DM 226/2021: Linguistic Activities

Teaching Activities ex DM 226/2021: Research management and Enhancement

Teaching Activities ex DM 226/2021: Statistics and Computer Sciences

Teaching Activities: Free choice

GENERALIZED LINEAR MODELS: LOGISTIC REGRESSION, LOGLINEAR MODEL, POISSON MODEL  (2023/2024)

Teacher

Lucia Cazzoletti

Referent

Lucia Cazzoletti

Credits

2

Language

English

Class attendance

Free Choice

Location

VERONA

MoodleMoodle Seminars 0

Learning objectives

An introduction to the fundamentals of generalized regression models will be given in this course, with a focus on models for count, binary, and categorical data. These types of response variables are widely used in industrial applications as well as observational and experimental research.
Upon successful completion of the course, students will be able to:
• Describe the general structure of a GLM and similarities and differences with linear models
• Estimate and interpret a logistic regression model
• Estimate and interpret a Poisson regression model
• Know of issues and some strategies for dealing with overdispersion in some generalised linear models (GLMs)

Prerequisites and basic notions

This course assumes a good understanding of probability and mid-level knowledge of linear regression theory.

Program

The course covers methods for regression analysis of responses that do not follow the normal distribution, especially of discrete responses. We will learn to understand some of the common statistical methods for fitting regression models to such data. In particular, we will consider logistic regression, Poisson regression and log-linear models. The lecture focuses on the development, theoretical justification, and interpretation of these methods.

Didactic methods

Teaching forms mainly consist of lectures (8h) and exercises proposed by the teacher. The teaching material (slides of the theoretical lessons) is made available to the students on the e-learning web page of the course (Moodle platform). Lessons will be delivered via Zoom. Full attendance is required.

Learning assessment procedures

There is no exam

Students with disabilities or specific learning disorders (SLD), who intend to request the adaptation of the exam, must follow the instructions given HERE

Assessment

There is no exam, hence there is also no definition of the evaluation criteria.

Criteria for the composition of the final grade

There is no grade because there is no exam.

Scheduled Lessons

When Classroom Teacher topics
Tuesday 19 March 2024
14:30 - 16:30
Duration: 2:00 AM 		https://univr.zoom.us/j/7951194173?pwd=ZVY2QzMzQ253QW9hSVFLY3d2M084dz09 Lucia Cazzoletti Introduction to generalised linear models. Review of the general linear model: assumptions of the linear model (independence of observations, homoskedasticity of errors, linearity of coeffiicents) , least squares estimation and maximum likelihood estimation. Main features of the genralised linear model: i) the probability distribution function of the random component of the response variable belongs to the exponential family, ii) a differentiable and monotonic link function relates the mean of the response variable to the linear predictor, a linear combination of coefficients and explanatory variables.
Tuesday 26 March 2024
14:30 - 16:30
Duration: 2:00 AM 		https://univr.zoom.us/j/7951194173?pwd=ZVY2QzMzQ253QW9hSVFLY3d2M084dz09 Lucia Cazzoletti Introduction to the theoretical basis of logistic regression model, commonly used for binary (proportion/percentage) data, as a generalised linear model. Binomial distribution for the outcome binary variable. The link between probability and logodds. Maximum likelihood estimation of the coefficients of the model. Interpretation of the meaning of the regression coefficients and their statistical significance.
Wednesday 03 April 2024
14:30 - 16:30
Duration: 2:00 AM 		https://univr.zoom.us/j/7951194173?pwd=ZVY2QzMzQ253QW9hSVFLY3d2M084dz09 Lucia Cazzoletti Introduction to the theoretical basis of Poisson regression model, commonly used for count data, as a generalised linear model. Poisson distribution for the outcome variable. Maximum likelihood estimation of the coefficients of the model. Interpretation of the meaning of the regression coefficients and their statistical significance. Use of the offset to take into account the different exposure of subjects. Extensions of the Poisson Regression Model: Negative binomial regression model (NBRM), Zero-inflated poisson (ZIP) model, Zero-truncated count data model.
Wednesday 24 April 2024
14:30 - 16:30
Duration: 2:00 AM 		https://univr.zoom.us/j/7951194173?pwd=ZVY2QzMzQ253QW9hSVFLY3d2M084dz09 Lucia Cazzoletti Using Deviances to Compare Models for Logistic and for Poisson Regression Models. Use of the Likelihood Ratio Test to assess the presence of overdispersion. Some hints about the log-linear model in the presence of contingency tables