Statistics for experimental design are performed to describe the results and to help clarify a conclusion giving a probability to accept or reject a hypothesis which is in many cases a hypothesis of no differentiation. For most bee study plans or protocols, the variables are mainly counting. Very few are issued from a quantified continue measure such as weight, length, etc. These measured variables can be mortality counts, foraging counts, behavioural counts such as toxicity signs or brood development, etc. These observed counts are raw data issued from experimenter observations in a laboratory box or cage, in a tunnel (semi-field condition), in a field, or directly in a hive. For these counts, two main situations are observed. In the first case, the size is exactly known as when a LD50 study is performed in cages with ten or twenty bees, or in a hive for a brood development study, 100 individual brood cells per hive are identified. In the second case, the size is not known. An estimation of population is made in the hive, and the counting is performed on the foraging activities or a counting of the dead bees is performed in the tunnel or in the field.
For most situations, several dose modalities are studied. The experimental design at a minimum includes a negative reference group as a sentinel to measure the experimental background noise (untreated or water treated control). A positive reference group is also often included to measure an experimental bias of no response (i.e. dimethoate). These two kinds of control permit one to validate (or invalidate) the study. Formal criteria are predefined in protocols.
An experimental test item modality is included in the experimental design. At least one modality is studied. The experimental design will include at minimum two or three groups, or product modalities, and up to ten or more product modalities. These modalities are usually independent. The same hive is not observed under several doses or product modalities but the hives are observed several times; then the counting is repeated. If the same modality is studied several times, replicates are observed and can be compared.