How to identify the experimental unit in an in vivo experiment.
- Why is the experimental unit important?
- Know your experimental unit
- Experiments with more than one experimental unit
- Representing the experimental unit in the EDA
The experimental unit is the entity you want to make inferences about (in the population) based on the sample (in your experiment).
The experimental unit is the entity subjected to an intervention independently of all other units. It must be possible to assign any two experimental units to different treatment groups.
The sample size is the number of experimental units per group. You need enough experimental units in your experiment for reliable results. But, if you do not correctly identify the experimental unit, there is a risk you overestimate your sample size which could invalidate the results of your statistical analysis and conclusions.
The British Pharmacological Society have created an animated video to introduce the concept of experimental units and how correctly identifying them is important to interpret your results.
In animal experiments the experimental unit is often the individual animal. In this case, each animal is allocated to a particular treatment group independently of other animals. But this is not always the case. Depending on the treatment administered, the experimental unit may be bigger than the animal (e.g. a litter or a cage) or smaller than the animal (e.g. part of the animal or an animal for a period of time). You can learn more about how to identify your experimental unit using the examples in this section.
Note that if you take multiple measurements from the same animal it does not mean that each animal provides multiple experimental units. The experimental unit is defined as the entity which receives an intervention or treatment, regardless of how many times you take measurements from it.
This is the most common situation and individual animals are independently assigned to distinct categories of the variable(s) of interest. It must be possible for any two individual animals to receive different treatments.
An example could be an experiment with four groups defined by two variables of interest, sex and exercise. The categories of the variables of interest are 'female with exercise', 'female no exercise', 'male with exercise', and 'male no exercise'. Animals are either male or female independently of other animals, and each animal is allocated to different activity levels independently of the other animals. Thus, the experimental unit is the individual animal.
Consider a teratogenesis experiment where the pregnant female receives a treatment and measurements are made on the individual pups after birth. Animals within a litter are all exposed to the same treatment – the experimental unit is therefore the whole litter. In this case, the variable ‘individual pups’ is nested into the experimental unit ‘litter’.
If animals are group housed in a cage and all animals within that cage receive the same treatment, for example in the drinking water or diet, then the experimental unit is the cage of animals.
However, if animals are group housed but can each receive a different treatment, for example by injection (and the treatment will not contaminate cage mates), then the experimental unit would be the individual animal.
If animals are exposed to a treatment via topical application, it may be possible to divide an area of skin into a number of different patches which can each receive distinct treatments. In this situation, the patch of skin on the animal is the experimental unit.
If individual cells can be stimulated independently and recording of the responses is made at the individual cell level, the experimental unit for the stimulation is the individual cell. Provided the experiment does not include another treatment which the whole animal is exposed to (e.g. drug injection or genotype), the individual cell can be the experimental unit for the whole experiment and a single animal provides many experimental units. It is important to note that if just a single animal is used, then the results hold true for that animal alone and cannot be generalised to the population.
When a single animal provides multiple experimental units, to avoid the confounding effect of between-animal variability, the individual animal should be used as a blocking factor and more than one animal should be used to improve generalisability. The number of animals needed depends on the between-animal variability.
Another scenario where a single animal can provide several experiment units is in a crossover experiment. In this experimental design, each animal is used as its own control and receives distinct treatments, separated by wash out periods. As animals can be exposed to different treatments in different test periods, the experimental unit is the animal for period of time.
Occasionally, there may be multiple experimental units in a single experiment, for example in a so-called split plot experiment.
Consider a situation where the effects of two different treatments (diet and vitamin supplements) on growth rate are investigated in mice. Diet is administered at the cage level and all mice housed in the same cage receive the same diet – the experimental unit for the diet treatment is therefore the cage. However, the vitamin supplement is administered by gavage meaning animals within the same cage can receive different supplements – the experimental unit for the vitamin supplement is the individual mouse.
This type of design is powerful as it enables researchers to investigate whether the effect of the vitamin is related to the diet administered. However the statistical analysis can be complicated and expert statistical advice should be sought before conducting such an experiment.
On your EDA diagram, the experimental unit is represented by the experimental unit node. This node is connected from one of the group nodes as shown in the image below.
If the experimental unit is the same throughout your experiment you only need one experimental unit node in your diagram. If there are multiple experimental units, multiple nodes may be necessary to clarify which unit different interventions are applied to.