GESIS Training Courses

Scientific Coordination

Alisa Remizova
Tel: +49 221 47694-475

Administrative Coordination

Noemi Hartung
Tel: +49 621 1246-211

Introduction to Quantile Regression

Online via Zoom
General Topics:
Course Level:
Software used:
Students: 220 €
Academics: 330 €
Commercial: 660 €
Additional links
Lecturer(s): Andreas Haupt, Sebastian E. Wenz

About the lecturer - Andreas Haupt

About the lecturer - Sebastian E. Wenz

Course description

Across different disciplines, social scientists typically turn to regression models that focus on conditional means-e.g., linear regression-for both descriptive and causal inference. However, instead of mean differences and average treatment effects (ATEs) we might be more interested in estimands for differences between, effects at, or effects on multiple other points of the outcome distribution-e.g., percentiles or quantiles in the lower and upper part of the distribution. This is where the increasingly used but often misunderstood quantile regression (QR) comes in: Depending on the exact estimator, QR addresses questions about…
  • distributional differences between groups (e.g., differences in the test-score distributions of immigrants and natives),
  • effects at certain points of the outcome distribution (e.g., whether the motherhood wage penalty differs between mothers with different wages),
  • or effects on the overall outcome distribution (e.g., the effect of a change in the minimum wage on the income distribution).
  • In this workshop, we will discuss the implementation and interpretation of different QR models and related estimators that correspond to all three perspectives mentioned above, namely:
  • Conditional Quantile Regression (CQR) and its estimator proposed by Koenker & Bassett (1978).
  • Quantile Treatment Effects (QTE), for which we will consider the estimators by Powell (2020) and Borgen et al. (2021).
  • Unconditional Quantile Regression (UQR) and the corresponding estimator proposed by Firpo et al. (2009).
  • Depending on the estimators, we discuss robustness to outliers, nonlinearity, interaction effects, and typical misconceptions and how to avoid them.
    An in-depth discussion of other QR estimators and QR models for longitudinal data or high-dimensional data is beyond the scope of this two-day workshop. If time permits, we will briefly cover some advanced QR techniques (e.g., decomposition techniques). However, after successfully participating in the workshop, participants will be well-equipped to engage with the advanced QR literature.

    Target group

    Participants will find the course useful if:
    • They are interested in research questions beyond the mean-be it differences between, effects at, or effects on multiple other points of the outcome distribution.
    • They are interested in learning when and how to implement different quantile regression (QR) estimators for different-conditional and unconditional-forms of QR.
    • They are interested in learning how to interpret and visualize estimation results of different QR estimators.

    Learning objectives

    By the end of the course participants will:
    • Know when and how to implement different quantile regression (QR) estimators.
    • Know how to interpret estimation results of different QR estimators.
    • Know how to visualize results of different QR estimators in Stata.


    • Solid knowledge of multiple (linear) regression from a course in an MA or PhD program is required.
    • Familiarity with Stata is highly recommended. We will rely on Stata for all exercises and practical examples.
    Software and hardware requirements
    For all exercises and practical examples, we will rely on Stata. Therefore, participants will need to bring a laptop computer with a recent version of Stata (13 or higher) installed to successfully participateparticipate successfully in this course. Stata short term licenses will be provided by GESIS for the duration of the course if needed.


    Recommended readings