Frank Konietschke, a Professor of statistics with expertise in non-parametric methods, dismantles the myth that non-parametric means just Wilcoxon tests. He explores a broad toolkit for analyzing skewed data, outliers, and small samples. Learn how ranks can quantify the relative treatment effect without relying on means, and discover effective ways to present results using confidence intervals and visuals. Frank also shares valuable software tools for implementing rank-based models, ensuring you don't miss the innovative strategies available for robust statistical analysis.
40:19
forum Ask episode
web_stories AI Snips
view_agenda Chapters
menu_book Books
auto_awesome Transcript
info_circle Episode notes
insights INSIGHT
What Non‑Parametric Means
Non-parametric means you do not postulate a specific data distribution like normality.
Ranking methods let you analyze metric, ordinal, and binary data without distributional assumptions.
volunteer_activism ADVICE
Prefer Ranks For Tiny Samples
Use rank-based methods for small sample studies where you cannot estimate the data distribution.
Prefer ranks for preclinical or animal studies with n like eight or nine per group.
insights INSIGHT
Robustness To Outliers
Rankings reduce the influence of extreme outliers because rank position, not magnitude, matters.
Rank methods are robust when outliers are far from the rest of the data.
Get the Snipd Podcast app to discover more snips from this episode
Why this episode made our all-time Top 9: If you’ve ever thought “non-parametric = Wilcoxon/Mann-Whitney and that’s it,” this conversation will happily destroy that myth. Frank shows how rank-based methods unlock rigorous analyses for skewed data, outliers, ordinal endpoints, small samples, composites/estimands—and how to communicate effects without relying on means.
You’ll walk away with:
✔ Non-parametric ≠ one test: A broad toolkit for two-group, multi-group, longitudinal, factorial, and covariate-adjusted designs.
✔ When ranks shine: Ordinal scales, heavy skew, small n (e.g., preclinical/animal studies), outliers, composite endpoints under the estimand framework.
✔ Interpretable effects without means: The probability-based “relative treatment effect”—“What’s the chance a random patient on A does better than a random patient on B?”
✔ Link to parametrics (when you must): How the rank-based effect relates to standardized mean differences under normality.
✔ Presenting results: Confidence intervals for rank-based effects and clean visualizations.
✔ Software exists: SAS macros and R packages for rank-based models (plus pointers to Frank’s book).
✔ Missing data & estimands: Practical thinking about composite strategies, treatment policy, and ongoing research for rank methods with missingness.
Episode Highlights:
00:00 – 03:31 | Welcome & setup TES resources, PSI community, and why innovative methods often struggle with adoption.
03:32 – 06:00 | Meet Frank From Göttingen to Munich, Texas, and back to Berlin; preclinical research focus.
06:01 – 09:11 | What are non-parametric analyses? No strict distributional model; works for metric, ordinal, and binary data.
09:12 – 12:13 | Why ranks? Small samples, unknown distributions; robustness when outliers occur.
12:14 – 14:35 | Where ranks are the better choice Ordinal ratings (A/B/C/… without meaningful distances), outliers, skew, composites.
14:36 – 21:18 | Defining the treatment effect without means Relative treatment effect as a probability (e.g., 60% = in 60% of random pairings, new treatment is better). Connection to parametric world under normality assumptions.
21:19 – 23:13 | How to present it Confidence intervals for rank-based effects and clear plots.
23:14 – 30:18 | Beyond two groups Multi-arm trials, repeated measures, factorial designs, covariate adjustments; pseudo-ranks and why unweighted references improve interpretability and power properties.
30:19 – 35:33 | Missing data, real-world setups & estimands Practical strategies (composites, treatment policy) and active research on rank methods with missingness.
35:34 – 39:41 | Collaboration & wrap-up Research networks, software, and how statisticians can lead method adoption.
References:
Book: Brunner, E., Bathke, A.C., Konietschke, F. (2019). Rank and Pseudo-Rank Procedures for Independent Observations in Factorial Designs -Using R and SAS. Springer
Brunner, E., Konietschke, F., Pauly, M., & Puri, M. L. (2017). Rank‐based procedures in factorial designs: hypotheses about non‐parametric treatment effects. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 79(5), 1463-1485.
Konietschke, F., Bathke, A. C., Hothorn, L. A., & Brunner, E. (2010). Testing and estimation of purely nonparametric effects in repeated measures designs. Computational Statistics & Data Analysis, 54(8), 1895-1905.
Konietschke, F., Hothorn, L. A., & Brunner, E. (2012). Rank-based multiple test procedures and simultaneous confidence intervals. Electronic Journal of Statistics, 6, 738-759.
Konietschke, F., Harrar, S. W., Lange, K., & Brunner, E. (2012). Ranking procedures for matched pairs with missing data—asymptotic theory and a small sample approximation. Computational Statistics & Data Analysis, 56(5), 1090-1102.
🔗 Medical Data Leaders Community – Join my network of statisticians and data leaders to enhance your influencing skills.
🔗 My New Book: How to Be an Effective Statistician - Volume 1– It’s packed with insights to help statisticians, data scientists, and quantitative professionals excel as leaders, collaborators, and change-makers in healthcare and medicine.
Join the Conversation: Did you find this episode helpful? Share it with your colleagues and let me know your thoughts! Connect with me on LinkedIn and be part of the discussion.
Subscribe & Stay Updated: Never miss an episode! Subscribe to The Effective Statistician on your favorite podcast platform and continue growing your influence as a statistician.
The Effective Statistician - in association with PSI 
