330SM - STATISTICAL LEARNING IN EPIDEMIOLOGY 2024
Section outline
-
-
Opened: Friday, 28 February 2025, 4:15 PM
Some basic information about your background and skills
-
This section will cover fundamental epidemiological concepts, including the distinction between association and causation, and common measures of disease frequency.
We will also review essential probability concepts and their practical application in epidemiological studies.
Finally, we will explore univariable statistical measures of association, both on the relative and absolute scale.
-
Epi intro File HTML
-
This section introduces fundamental study design principles, including sampling from a target population and the basic ideas about RCT (randomized controlled trials) and observational studies.
Subsequently, we will explore core sample size determination rules (from a frequentist perspective), applicable to both RCTs and observational studies.
Theoretical concepts are illustrated through slides, and practical R code examples are provided for hands-on application.
-
Block 2.1 File PDF
-
Block 2.2 File PDF
-
Block 2.3 File PDF
-
Block 2_4 File PDF
-
-
In block 3, we will focus essentially on multivariable regression models.
We will present the 3 uses of these models in clinical and epidemiological research: descriptive, predictive, and causal models.
Of note : often descriptive models are just the first step in the statistical analysis, then according to the main scientific question, the model is targeted toward a predictive or causal aim.
We will also provide some guidance on performance evaluations in the context of predictive models and on sample size in a multivariable setting.
-
In Block 4 we focus on survival analysis, where time to event is the primary outcome. We will introduce core survival analysis techniques, beginning with the intuitive Kaplan-Meier (KM) curve, a non-parametric method for visualizing and estimating the survival probability over time.
Then, we will delve into the Cox Proportional Hazards model, a powerful and widely used semi-parametric regression technique that allows us to investigate the influence of multiple covariates on the hazard rate.
Furthermore, we will introduce the concept of competing risks, situations where individuals are at risk of experiencing "first" multiple distinct events, and understanding how to properly analyze such data is crucial for obtaining unbiased estimates.
Finally, we will touch upon the performance evaluation of estimated survival models, exploring metrics and approaches to assess how well our models fit the data and how reliably they can predict future event times.
-
-