Explanatory modeling of observational quantitative data: Causal graphs, interactions, and research design choices (2025)

Welcome to Explanatory modeling of observational quantitative data

"Explanatory modeling of observational quantitative data" is an applied course designed for PhD students in quantitative social sciences who wish to deepen their understanding and skills in working with observational quantitative data (e.g., non-experimental data). Causality is hard to establish with observational data, yet a theory-testing approach still forces researchers to formulate causal theoretical models to motivate their statistical choices. This course works with this tension in quantitative social science research and engages with applied recommendations for best practice. 

The first half-day of the course introduces directed acyclic graphs and discusses the need for causal reasoning for regression-based modeling.

The second half of day one is dedicated to further complexities emerging from interactive and non-linear hypotheses, introducing statistical packages to ensure valid inferences for these statistical scenarios.

The second day introduces ‘causal’ research designs to observational data, focusing on matching estimators and synthetic control methods. We compare these to regression-based approaches and discuss strengths and weaknesses. We end the second day with a Lab session that provides the chance to apply some of the content to your own research.   

The course expects a basic familiarity with quantitative methods (e.g. linear regression). The applied statistical teaching is done with R and students are recommended to have basic knowledge of R programming. Most of the course content, however, can also be followed with STATA (e.g, similar/same packages in STATA). Students are encouraged to bring their own research questions to the course and engage with potential inferential/modeling challenges in their field during the practical parts of the course.

Through a combination of lectures, practical exercises, and case studies, you will engage with current best practices in observational quantitative social science research. You will learn about the use of directed acyclic graphs. You will explore various statistical tools such as kernel or bin plots to assess the validity of linear interactive models. We will also touch upon more advanced machine-learning-based models, such as the Kernel Regularized Least Squares estimator, to uncover non-linear patterns in your data. Finally, you will be enabled to make informed choices of whether matching or synthetic control methods might be useful tools for your analysis.

By the end of the course, you will have a comprehensive toolkit of advanced statistical approaches to tackle complex research questions in quantitative social sciences. You will also gain the ability to critically evaluate existing literature and design rigorous empirical studies.

Teaching methods:

• Lectures, practical exercises, and case studies
• Applied programming with R
• Illustration of the statistical approaches with real world cases and data
• Opportunities to work with your own data during the course

Programme outline:

1^st day

08:00-11:00 Directed acyclic graphs for regression-based analysis (lecture)

12:00-16:00 Interactive and non-linear hypotheses

2^nd day

08:00-10:00 ‘Causal’ designs to observational data: Matching (lecture)

10:00-14:00: Presentations

14:00-15:00 ‘Causal’ designs to observational data: Synthetic control method (lecture)

15:00- 16:00 Questions

3^rd day

08:00-09:00 Recap and Lab Session (lecture)

09:00-12:00 Questions

Description of paper requirements:

The final paper is a post-reflection paper that be uploaded. The paper should present the whole workflow of (1) a hypothesis, (2) acquiring and preparing the data to test it, (3) specifying and justifying the statistical model to test the hypothesis, and (4) assess the statistical validity of the findings.

The emphasis should lie on the last point (4), so students are encouraged to use readily available data, e.g. replicating existing studies proposing interactive relationships.

Organizer:

Dominik Schraff, Department of Politics and Society, Political Sociology

Lecturers:

Dominik Schraff, Department of Politics and Society, Political Sociology

ECTS:
3

Time:
21, 22, 23 May 2025

Place:
Kroghsstræde 3, Room 5.130 - Finsen

Zip code:
9220

City:
Aalborg

Number of seats:
15

Deadline:
30 April 20225

30 June 2025 - upload in the assignment below

Key literature:

Spirling, A. & Stewart, B. M. (2024). What Good is a Regression? Inference to the Best Explanation and the Practice of Political Science Research. Journal of Politics, forthcoming.

Keele, L., Stevenson, R.T. and Elwert, F. (2020) The causal interpretation of estimated associations in regression models. Political Science Research and Methods, 8(1), pp. 1–13.

Hainmueller, J., Mummolo, J., & Xu, Y. (2019). How Much Should We Trust Estimates from Multiplicative Interaction Models? Simple Tools to Improve Empirical Practice. Political Analysis, 27(2), 163-192.

Xu, Y. (2017) Generalized Synthetic Control Method: Causal Inference with Interactive Fixed Effects Models. Political Analysis, 25(1), pp. 57–76.

Mandatory literature:

Spirling, A. & Stewart, B. M. (2024). What Good is a Regression? Inference to the Best Explanation and the Practice of Political Science Research. Journal of Politics, forthcoming.

 Keele, L., Stevenson, R.T. and Elwert, F. (2020) ‘The causal interpretation of estimated associations in regression models’, Political Science Research and Methods, 8(1), pp. 1–13. doi:10.1017/psrm.2019.31.

Hainmueller, J., Mummolo, J., & Xu, Y. (2019). How Much Should We Trust Estimates from Multiplicative Interaction Models? Simple Tools to Improve Empirical Practice. Political Analysis, 27(2), 163-192. doi:10.1017/pan.2018.46

Giesselmann, M., & Schmidt-Catran, A. W. (2022). Interactions in Fixed Effects Regression Models. Sociological Methods & Research, 51(3), 1100–1127. https://doi.org/10.1177/0049124120914934

Beiser-McGrath, J., & Beiser-McGrath, L. (2020). Problems with products? Control strategies for models with interaction and quadratic effects. Political Science Research and Methods, 8(4), 707-730. doi:10.1017/psrm.2020.17

Brambor, T., Clark, W. R., & Golder, M. (2006). Understanding Interaction Models: Improving Empirical Analyses. Political Analysis, 14(1), 63–82. http://www.jstor.org/stable/25791835

Imai, K., Kim, I.S. and Wang, E.H. (2023), Matching Methods for Causal Inference with Time-Series Cross-Sectional Data. American Journal of Political Science, 67: 587-605. https://doi.org/10.1111/ajps.12685

Xu, Y. (2017) Generalized Synthetic Control Method: Causal Inference with Interactive Fixed Effects Models. Political Analysis, 25(1), pp. 57–76.

Important information concerning PhD courses: 

There is a no-show fee of DKK 3,000 for each course where the student does not show up. Cancellations are accepted no later than 2 weeks before the start of the course. Registered illness is of course an acceptable reason for not showing up on those days. Furthermore, all courses open for registration approximately four months before start of the course.

We cannot ensure any seats before the deadline for enrolment, all participants will be informed after the deadline, approximately 3 weeks before the start of the course.

For inquiries regarding registration, cancellation or waiting list, please contact the PhD administration at phdcourses@adm.aau.dk When contacting us please state the course title and course period. Thank you.