3. Experimental design

There are five components to an experiment: hypothesis, experimental design, execution of the experiment, statistical analysis, and interpretation (Hurlbert, 1984). To be able to analyse data in an appropriate manner, it is important to consider one’s statistical analyses at the experimental design stage before data collection, a point which cannot be emphasised enough.

Critical features of experimental design include: controls, replication and randomisation; the latter two components will be dealt with in the next section (3.1.). In terms of a ‘control’ in an experiment: a negative control group is a standard against which one contrasts treatment effects (untreated or sham-treated control), whereas a positive control group is also often included usually as a “standard” with an established effect (i.e. dimethoate in the case of toxicological studies, see the BEEBOOK paper on toxicology by Medrzycki et al., 2013). Additionally, experiments conducted blind or double blind avoid biases from the experimenter or observer. If that is not possible, one should control for the biases of observers by randomly assigning several different observers to different experimental units or by comparing results from one observer with previous observers to quantify the bias so one can account for it statistically when interpreting results of analyses.

3.2. Sample size determination