Randomization Modelling of the Crossover Experiment for Clinical Trials

O. Gomez‐Marin, R. B. McHugh

Research output: Contribution to journalArticlepeer-review

3 Scopus citations


The two‐period cross‐over experiment for clinical trials has been examined by several writers following a Gaussian linear model approach. Some authors have expressed interest in the “derivation of the finite permutation model” and have pointed out that the randomization approach to modeling the two‐period cross‐over design “would highlight the importance of randomizing the subjects to the two groups as a basis for inference”. However, in the literature, there is no development of the randomization approach to this important design. In this paper, after a statement of the experimental design and formulation of the observation random variables of the finite population, two additive randomization models—one with residual effects, the other without—which are the analogues of Grizzle's Gaussian models, are derived. Statistical inference is developed for these randomization models and the results are compared with those of the corresponding Gaussian models. Also, exact inference based upon Fischer's approach is presented.

Original languageEnglish (US)
Pages (from-to)901-914
Number of pages14
JournalBiometrical Journal
Issue number8
StatePublished - 1984


  • Fischer's exact test
  • Gaussian linear model
  • Randomization model
  • Two‐period cross‐over design

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty


Dive into the research topics of 'Randomization Modelling of the Crossover Experiment for Clinical Trials'. Together they form a unique fingerprint.

Cite this