ABSTRACT

What considerations should guide us when we are planning an experiment? At the simplest level, there are two basic statistical requirements for a good experiment: replication and randomization.

The need for replicating-each treatment being applied to more than one experimental unit-should by now be clear. The material to be used in the experiment will always be variable. To have any idea of the “value” of the results of an experiment and of the validity of the conclusions drawn from the results of an experiment, we must obtain an estimate of this variability. This can only be done with a sample of results for each treatment. In addition, the more units used for each treatment the lower will be the standard errors (SEs) of the estimates of treatment effects and, therefore, the more accurate the experiment.