Notes on designing an experiment

What are the treatments?

Here you need to give a precise description of the treatments that you intend to apply to the experimental units. Give complete technical details, such as

5mg of ciprofloxacin 4 hours after contact.

Then give each treatment a simple code like A, B, C for reference later.

It is usually helpful to state how many treatments there are as well as what they are. In the bees example, there are 3 treatments.

Sometime treatments are simple; sometimes they are combinations. In the ciprofloxacin example, there may be 4 different doses combined with 2 times of administration: this would give 8 treatments. If all doses are administered at 4 hours then the information about `4 hours' belongs in Methods rather than here. Likewise, if all doses are the same and the purpose of the experiment is to find the best time to administer the drug, then the treatments are just times of administration and all details about dose and drug go into Method.

In the waking example, we should ask the professor if he just wants to compare the two new pills with each other or whether he wants to compare them both with the effect of doing nothing. If the latter then there is a third treatment, `do nothing', which is often called control. You should always ask if a control is needed. Scientific orthodoxy says yes, but there are experiments where a control can be harmful. If there is already an effective therapy for a disease then it is unethical to run an experiment comparing a new therapy to `do nothing'; in this case the treatments should be the new therapy and the one currently in use. In a trial of several pesticides in one field, if there is a `do nothing' treatment on some plots then the pest may multiply on those plots and then spread to tbe others.

In experiments on people, `do nothing' should often be replaced by a placebo, so that everyone involved thinks that something is being done.

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