In This Story
Statistics Professor William F. Rosenberger believes that good statisticians, like good detectives, should seek a randomization process that works for the specific goals of their planned trial. That is why his research focuses on the randomization of clinical trials: how to randomize, when randomization is appropriate, and how to discover new ways of randomization. When Rosenberger was asked to present a lecture for the Fisher Memorial Trust at Cambridge University as the 41st Fisher Memorial Lecturer, he jumped at the chance.
“Sir Ronald Fisher, for whom the lecture is named, was arguably our most famous statistician, and one might call him the founder of a number of the techniques that we use regularly in statistics,” said Rosenberger. “Much of what we do in experimental design is based on his famous book, The Design of Experiments, and much of what we do in applied statistics, is based on his book Statistical Methods for Research Workers. The books were written back in the 1930s and 40s and we still use them today.”
Rosenberger’s soon-to-be-published article in the Journal of the Royal Statistical Society (JRSS) was a direct outcome of his appearance at Cambridge University last year. The Fisher Memorial Trust comprises a group of distinguished British statisticians and most of them attend the lecture. One of them, Professor Stephen Senn, approached Rosenberger about submitting a paper to JRSS based on his lecture for publication.
The article, From Fisher to CARA: The Evolution of Randomization and Randomization-based Inference, shows a different side of Fisher—the less well-known, more irascible side. “We look at Fisher now and everybody thinks of the great Fisher,” said Rosenberger. “Well, he wasn’t always great. His career had ups and downs, he succeeded, but he also failed,” said Rosenberger.
Though Rosenberger sheds new light on Fisher’s career, the focus of the talk and the the article is the history of randomization. The article explores if is there a way to modify randomization so, that in clinical trials, more patients get better treatments, which has been a major part of Rosenberger’s research for many years.
The other piece examines how to analyze an experiment that's been randomized. “Fisher told us how to do that in his Design of Experiments book. No one knew how to do that before. And it's the basis of randomization tests. Which he invented. And they're not used very often today, and I try to explain why they're not used very often. And why they should be used more often,” said Rosenberger.
In the article’s conclusion, Rosenberger encourages statisticians and biostatisticians to “leverage randomization to their advantage to uncover logistical, statistical, and ethical goals.”
To read the full article go to https://academic.oup.com/jrsssa/advance-article/doi/10.1093/jrsssa/qnaf002/8029788