Related Experiment Videos
Controversies concerning randomization and additivity in clinical trials
1Department of Statistical Science, University College London, London WC1E 6BT, UK. stephen@senns.demon.co.uk
Statistics in Medicine
|December 8, 2004
Summary
Randomization in clinical trials is common, but analysis often deviates from the randomization method. This paper examines historical debates and modern relevance of randomization inference versus modeling.
Area of Science:
- Statistics
- Clinical Trials
- Experimental Design
Background:
- Fisher's principle 'as ye randomise so shall ye analyse' is often cited but not always followed in clinical trials.
- Historical controversies involving Fisher, Neyman, and Student highlight debates on randomization and analysis methods.
Observation:
- Clinical trials frequently employ randomization, yet the analysis methods used do not always align with the randomization scheme.
- Fisher's own work demonstrated deviations from strict adherence to randomization in analysis.
Findings:
- The paper explores the ongoing relevance of randomization inference versus statistical modeling in clinical trials.
- It questions the acceptability of minimization procedures and the confusion between experimental and survey designs.
Implications:
- Conflicting purposes of clinical trials may arise from the conflation of analysis methods, particularly with the general linear model.
- Understanding the nuances of randomization and analysis is crucial for appropriate interpretation of clinical trial results.