Causal Analysis Without Experiments

Causal analysis is a central approach in research that explores why a phenomenon occurs or how it comes about. Randomized controlled trials (RCTs) are the gold standard for investigating cause and effect: individuals are randomly assigned to two groups that are identical in every respect except for one examined factor — the well-known medication vs. placebo setting. If the groups differ in their outcomes (e.g., one group remains ill while the other recovers), we can attribute this difference to that factor, since no other systematic differences exist between the groups.

However, RCTs are generally not available in the Digital Humanities, as experiments cannot be conducted: historical actors are no longer accessible, texts have already been written, past events cannot be repeated, and societies cannot be experimentally manipulated. We are therefore reliant on observational data. How can we nevertheless conduct causal analyses in such contexts and answer questions such as: How did the European Reformation affect economic growth in 16th-century Europe? How does the gender of authors influence the stylistic features of their novels? And why does the diversity of genres in a theatre's repertoire affect its audience numbers?

In this workshop, we will explore methods of causal inference for observational data. First, we will deepen our theoretical understanding of causality using the Potential Outcomes Framework and use visual tools to construct causal models — that is, to make the assumed relationships between variables in a concrete research question explicit. In a second step, we will gain an overview of established statistical methods of causal inference, including Regression Discontinuity Designs, Instrumental Variables, and Difference-in-Differences. In a third step, we will apply one of these methods — Difference-in-Differences — to our own datasets.

The workshop is aimed at anyone who investigates causal questions in their research and is looking for a systematic approach.

It is also suitable for participants who do not yet have their own dataset but wish to apply causal inference in the future: participants can work with a sample dataset and learn what requirements their future data must meet for causal analyses, which will facilitate data generation.

To prepare for the BYODL, you are welcome to watch the following self-study unit on inferential statistics by Dr. Golnaz Sarkar Farshi: https://hermes-hub.de/lernen/resourcebase/resources/data-visualization-for-storytelling-and-statistical-inference.html

About the Expert:

Dr. Ramona Roller is a postdoctoral researcher in Computational Sociology at the Department of Sociology at the Universities of Groningen and Utrecht in the Netherlands.

Her research sheds light on social transformations and cooperative behaviour in modern and historical groups, such as work teams in the IT industry and scholarly communities during the European Reformation. For example, she analyses how hierarchies within a group affect its productivity, or how Protestantism spread across Europe. To do so, she uses large data collections (e.g., letter corpora and collaboration platforms), network analyses, and causal inference methods.

Registration

Registration is open until June 26, 2026. The number of participants is limited to 10.

Please use the following form: https://nocodb.nfdi4culture.de/dashboard/#/nc/form/f5cda303-05e4-444b-9a98-310847faa46c

Contact

For any questions, please contact: 

Johanna Konstanciak: konstanciakuni-trier.de (konstanciak[at]uni-trier[dot]de)

Preliminary Programme

• 9:15 a.m. - 9:30 a.m. – Welcome & introductory remarks
• 9:30 a.m. - 10:15 a.m. – Part 1a: Deepening the theoretical understanding of causality
• 10:15 a.m. - 11:05 a.m. – Part 1b: Applying the theoretical tools to one's own research question
• 11:05 a.m. - 11:15 a.m. Break
• 11:15 a.m. - 11:30 a.m. – Part 1c: Presenting the application results in plenary
• 11:30 a.m. - 12:15 p.m. – Part 2: Overview of statistical methods of causal inference
• 12:15 p.m. - 13:15 p.m. – Lunch break
• 13:15 p.m. - 13:30 p.m. – Part 3a: Introduction to the practical section
• 13:30 p.m. - 15:15 p.m. – Part 3b: Applying statistical methods to own datasets
• 15:15 p.m. - 15:30 p.m. – Part 3c: Presenting the application results in plenary
• 15:30 p.m. - 15:45 p.m. – Feedback round & closing remarks