Notes on designing an experiment

What are the experimental units?

The experimental units are what you apply the treatments to. Describe them carefully. It is helpful to note how many there are. For example (the dots denote missing technical details)

  • 20 male Wistar rats ...
  • 3 parts of the orchard in the bees example (which parts? how separated? what area?)
  • 12 student volunteers each used every day for 5 days
    (this is 60 experiemental units if treatments can be changed daily)
  • 11 female students aged 19-20 in the waking example.

    How should the experimental units be chosen? Should they be as alike as possible or be representative? If they are unalike then any variation between the experimental units may swamp any differences between the treatments that you are trying to find, especially in a small experiment. If they are not representative, you cannot extrapolate your conclusions.

    The waking experiment is small. I suggest ignoring the older students and then choosing students all of the same sex. If you find that the two pills affect young student women differently then you might plan further experiments with men and women, students and non-students, of various ages. In either case, the choice of alike experimental units or representative experimental units should be deliberate. Do not choose them at random.

    protocol purpose treatments experimental units method observational units
    measurements design justification randomization plan analysis

    Page maintained by R. A. Bailey

    Page updated 17/11/01