Experimental planning and statistical methodology

One of Professor Walløe's many fields of interest is statistics. Since the mid-1970s, he has advised medical researchers on the planning and analysis of controlled clinical trials. For several years, he was a member of a committee for clinical trials at the Norwegian Medicines Control Authority.

In addition to applied research, Professor Walløe's interests extend to the theoretical properties of statistical methods. He has focused particularly on the importance of using robust statistical methods, i.e. methods that are not overly sensitive to deviations from the assumptions on which a model is based, for example deviations from normal distribution. He has supervised several graduate students who have studied the robustness of statistical methods using stochastic simulation. This is a much more appropriate method of studying how a statistical method functions in practice than studying its asymptotic properties. A medical experiment usually only involves the collection of a relatively small number of observations, and stochastic simulation, in contrast to asymptotic calculations, makes it possible to study the properties of a method in realistic situations with a small number of observations.

Professor Walløe has also contributed to methodological developments in another field of statistics, sequential methods. The main objective of sequential experiments is to reduce the number of patients it is necessary to include in a clinical trial. Sequential methods differ from traditional methods in that a significance test is run each time the response of a new patient is registered. To determine when enough information has been collected for a trial, specially developed stopping rules are used. Professor Walløe's idea was to use a robust test as a basis and develop stopping rules by stochastic simulation.

In addition to supervising research students working on statistical methods and applied statistics, Professor Walløe has been teaching statistics to medical and pharmaceutical students for many years and has a reputation as a good lecturer. He was co-author of the textbook Elementær statistikk (Elementary statistics), first published in 1977, together with Professor Arnljot Høyland. Unlike most other textbooks at this level, it focuses on non-parametric methods.

Professor Eva Skovlund,
Department of Clinical Cancer Research
Norwegian Radium Hospital