gms | German Medical Science

66. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS), 12. Jahreskongress der Technologie- und Methodenplattform für die vernetzte medizinische Forschung e. V. (TMF)

26. - 30.09.2021, online

Comparing strategies for analysing count data in randomized controlled studies with pre- and post-tests – a simulation study

Meeting Abstract

Suche in Medline nach

  • Sebastian Appelbaum - Otto-Friedrich-Universität Bamberg, Bamberg, Germany
  • Thomas Ostermann - Universität Witten/Herdecke, Witten, Germany
  • Uwe Konerding - Otto-Friedrich-Universität Bamberg, Bamberg, Germany

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. 66. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS), 12. Jahreskongress der Technologie- und Methodenplattform für die vernetzte medizinische Forschung e.V. (TMF). sine loco [digital], 26.-30.09.2021. Düsseldorf: German Medical Science GMS Publishing House; 2021. DocAbstr. 77

doi: 10.3205/21gmds094, urn:nbn:de:0183-21gmds0942

Veröffentlicht: 24. September 2021

© 2021 Appelbaum et al.
Dieser Artikel ist ein Open-Access-Artikel und steht unter den Lizenzbedingungen der Creative Commons Attribution 4.0 License (Namensnennung). Lizenz-Angaben siehe



Very often outcome variables in health services research consist of count data. Although they are usually not symmetrically distributed, data of this kind are very often analyzed using statistical tests presupposing normal distributions. An alternative is using Poisson regression models, which are specially designed for count data. The aim of this simulation study is to compare six analysis strategies, 1) t-test, 2) Poisson regression model with MLE, 3) Poisson regression model with robust estimators 4) t-test for change scores, 5) Poisson regression with MLE with pre-test data as covariates, and 6) Poisson regression with robust estimators with pre-test data as covariates. Simulation data were generated using three different sample sizes, four average count number at pre-test, four different effect sizes and four pre post correlations. The study produced two main results: 1) the actual type 1 error probability for the MLE-Poisson regression with pre-tests as covariates does not agree with the significance level; 2) the robust Poisson regression with pre-tests as covariates seems to be the best test-strategy.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.


Hur K, Hedeker D, Henderson W, Khuri S, Daley J. Modeling clustered count data with excess zeros in health care outcomes research. Health Services and Outcomes Research Methodology. 2002;3(1):5-20.
Coxe S, West SG, Aiken LS. The analysis of count data: A gentle introduction to Poisson regression and its alternatives. Journal of personality assessment. 2009;91(2):121-136.
Nussbaum EM, Elsadat S, Khago AH. Best practices in analyzing count data: Poisson regression. Best practices in quantitative methods. 2008:306-323.
Zeileis A. Object-Oriented Computation of Sandwich Estimators. Journal of Statistical Software. 2006:16(9):1–16. DOI: 10.18637/jss.v016.i09 Externer Link
O'Connell NS, Dai L, Jiang Y, Speiser JL, Ward R, Wei W. Gebregziabher M. Methods for analysis of pre-post data in clinical research: a comparison of five common methods. Journal of biometrics & biostatistics. 2017;8(1):1-8.
Li H, Demirtas H, Chen R. RNGforGPD: An R Package for Generation of Univariate and Multivariate Generalized Poisson Data. R JOURNAL. 2020;12(2):173-188.