Biology, images, analysis, design...
Use/Abuse Principles How To Related
"It has long been an axiom of mine that the little things are infinitely the most important" (Sherlock Holmes)

Search this site




Unequal Randomization

The normal approach in any experiment is to have similar numbers of subjects in the different treatment groups. This is because this is the most efficient way to compare treatment means. However, one can argue that much information is already available on the control (whether receiving a placebo or the 'standard' treatment), whilst information is urgently needed on a new treatment. This might justify putting more experimental units in the treatment than in the control group. The same might be argued ethically in a clinical trial if there is good reason to believe that the new treatment will save lives.

In fact the efficiency of the comparison only declines markedly if more than about 75% of the patients are on the new treatment. Hence it may be acceptable to have up to three times as many patients in the treatment group compared to the control group. Note, however, there will be some loss in your ability to show significant differences between the groups.

An alternative approach is to allocate to 4 equal-sized treatment groups, of which one is a control, and three receive minor variations of the same treatment. That way you can test their combined effect against the control, and which (if any) variant is best. Or, if the treatments differ in dose rate, you can see whether there is any obvious trend in their outcome.