Example 4

Effect of exercise on neuronal density

View experiment diagram | Use this diagram as a template

An experiment is conducted to test the hypothesis that exercise has an effect on neuronal density in the hippocampus. To test this hypothesis, ten female and ten male mice are randomised into two different levels of activity: no running, or running for 30 min per day. After three weeks, mice are euthanized and histological brain sections are prepared; four slides are prepared from the hippocampus of each mouse. Neuronal density is assessed by counting the number of neurons on each slide.

Five mice of each sex are randomly allocated to each of the two treatment groups; this ensures that the groups are balanced with respect to treatment and sex. The allocation sequence is generated within the EDA and emailed to a colleague who will code the histological slides. The treatment allocation cannot be concealed as the investigator can see whether the wheel in the test cage can turn but the measurement is carried out blind and each slide is individually coded so that the animal they came from and the treatment group cannot be identified. 

In this experiment there are two variables of interest: Exercise, which has two categories: no running and running, and sex, which also has two categories: male and female to test the main hypothesis that exercise affects hippocampal neuron density and the secondary hypothesis that this varies between males and females. As the four histological slides were randomly selected from each mouse and we are not interested in the effect of the slide number, this is considered a nuisance variable and it is nested within the animal as each set of four slides are specific to each particular animal. This variable is not included in the analysis, as the neuron counts for each slide are averaged for each animal; thus a summary measure per animal rather than the raw values are analysed. If the data fits parametric assumptions, it can be analysed with a factorial ANOVA with two factors of interest (2 way ANOVA with interaction). The analysis is also carried out blind, the data is organised by the person who coded the slides into ‘Group A’ or ‘Group B’, so that the person doing the analysis knows which slides belongs to the same animal and which animals are grouped together but not what treatment they animals received.


Randomisation with sex as a factor of interest | Nuisance variable | Nested variable | Two-way ANOVA with interaction | Data averaged per animal | Multiple animal characteristics


This experiment is loosely based on example 3.8 (Bate and Clark, 2014)

Bate, ST and Clark, RA (2014). The Design and Statistical Analysis of Animal Experiments. Cambridge University Press.

First published 07 April 2014
Last updated 20 September 2023