Draw your research hypotheses using causal diagrams.

Date:

June 25 to 28 . AFTERNOON: 15 to 19h (June 25, 26 and 27) and 15 to 18h (June 28).

Instructor

Albert Sánchez Niubo 

Albert Sánchez Niubò holds a PhD in Biostatistics from the Faculty of Mathematics at the University of Barcelona (2014). He is currently professor in the Department of Social Psychology and Quantitative Psychology at the University of Barcelona.

He has been a researcher and methodological technician in the public health and epidemiology research context: at the Mar Institute for Medical Research (IMIM) (2003-2014) and Sant Joan de Déu Research Institute (2015-2021). He has been associate professor (part-time) at the Faculty of Medicine of the UB (2010-2014), at the UIC (2019-2021), and visiting professor at the UPF (2014-2015).

https://orcid.org/0000-0003-0309-181X

Language

Spanish. Slides in English

Description

Causal diagrams are graphical representations of causal relationships among variables in a specific research context. They have become increasingly popular in many fields, including health and social sciences, as they provide a powerful tool for investigating causal relationships and predicting the impact of interventions.

Causal diagrams can help researchers identify confounding variables, test causal hypotheses, and visually represent complex causal relationships. Understanding these relationships can help to improve treatment strategies, elucidate biases in observational studies and, consequently, inform more adequate policy decisions.

This course will explore the different types of causal diagrams, their components, and how to use them.

We will also discuss the advantages and limitations of using causal diagrams to understand causal relationships. Overall, causal diagrams are an essential tool for anyone interested in understanding and testing causal hypotheses.

Course goals

1. Introducing participants to the concept of causal diagrams and their role in investigating causal relationships among variables in a specific research context.
2. Teaching participants how to draw and use causal diagrams as a useful tool for identifying confounding variables in observational studies.
3. Providing participants with an understanding of the different types of causal diagrams and their components.
4. Demonstrating how to create and validate causal diagrams in practice.
5. Highlighting the advantages and limitations of using causal diagrams to understand complex causal relationships in research.
6. Equipping participants with the skills needed to test causal hypotheses using causal diagrams. Overall, the course aims to provide participants with a comprehensive understanding of effectively using causal diagrams in research to identify causal relationships, mitigate biases, and improve decisions based on data.

Course contents

1. Introduction

• Definition of causality
• Historical review
• The causal revolution

2. Counterfactual causality

• Definition of causal effects
• Causal effects in randomized experiments
• Causal effects in observational studies

3. Causal diagrams

• Directed Acyclic Graphs (DAGs)
• Unconditional association
• Conditional independence

4. Theory of causal DAGs

• D-separation rules
• The Adjustment Formula
• The Backdoor criterion
• The Front-door criterion

5. Unveiling paradoxes with DAGs

• Simpson’s paradox
• Lord’s paradox
• Berkson paradox
• Monty Hall paradox

6. Some practical applications

• The table 2 fallacy
• Inverse probability weighting
• Basics on Marginal structural models.

Prerequisites

Basic knowledge of statistical modelling and experience of having been actively involved in a research project can help to understand and appreciate the practical use of causal diagrams.

Targeted at

Students, academics and researchers especially in the fields of health and social sciences.

Evaluation

During the course, there will be in-class many exercises to help participants understand how causal diagrams work and how to use them in their research. The last day there will be time to make a final test.

Software requirements

Most of the exercises will be paper and pencil based. However, at some point, we will use some databases and be required to make simple regression models to test causal diagrams. For this, it is recommended to bring a laptop with some statistical software such as R, R-commander, Jamovi, or any other statistical software.