Notes on designing an experiment

Proposed statistical analysis

Write down guidelines for the statistical analysis that you propose to do once the data are collected. Do this before you collect the data: if you cannot think how to analyse the data then it is a waste of effort to collect it.

For a simple experiment with two treatments and no blocks, you would probably write `two-sample t-test'. For three treatments and no blocks, put `analysis of variance with three treatments and no blocks'.

Do not wait till you have the data before verifying that you can do the necessary statistical calculation, either with a calculator or using appropriate software. Unless you have recently analysed an experiment of exactly the same type, it is a good idea to invent some dummy data (or copy some from an old experiment) and check that you know how to get your software to do the correct analysis.

Of course, you may end up doing a different analysis from the one you proposed. This can happen if somethng goes wrong during the experiment, or if further information comes to light.

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