3. The experimental unit

Definition:

“The smallest division of the experimental material such that any two experimental units can receive different treatments”.

 It is the unit of randomisation and of statistical analysis when comparing groups

expU1In this study the animals are all housed in one cage and the treatment is given by injection.

Any two animals can receive different treatments, so the animal is the experimental unit and “N” (the total number of subjects) is 8

 

expU2

In this study the animals are housed two per cage and the treatment is given in the food or water.

What do you think is “N”, the total number of experimental units in this case?

Four
Two
Eight
Yes, each cage is an
experimental unit and
there are four of them

exp3b 

The experiment on the left has seven fish in each of two tanks. The left-hand tank has been treated with a test substance poured into the water and the right-hand has only the vehicle as a control. The aim is to measure the level of an enzyme in the fish.

What is the total number of experimental units?

 

Yes, there are only two experimental
units. Fish in a tank can not receive
different treatments so they are not
the units
Two
Seven Fourteen

In a crossover experiment an animal could be given a treatment for a period, then rested and given a different treatment for a period.  It is assumed that the treatment doesn’t alter the animal, so it has to be very mild.

expU302

In this experiment animals are given four treatments, sequentially in random order.  What do you think is N, the total number of experimental units?   

Three
Yes, the different time periods 
can receive different
treatments so the experimental
unit is an animal for a period of time.
In the statistical analysis account 
will need to be taken of
differences between animals
Four
Twelve

A within-animal experiment could be replicated in space. One eye could be a control for the other one. Patches on the shaved back could be used to test topical applications of a test substance. Statistical analysis of the results of such a within-animal should take account of differences between animals. This is one form of a “randomised block” design described later.

 In a teratology experiment (right) the pregnant femaleexpU5b is treated with the test compound or a placebo. The pregnant females are killed at about mid-gestation and the pups are weighed, measured and studied for abnormalities.

How many experimental units are there in the experiment on the right (two females, three pups from each)

Yes, the pregnant females is the
experimental unit because she
was assigned to a treatment, 
and each pregnant rat can receive
a different treatment
Two
Three
Six

Humans who suffer from depression seem to be more sensitive to pain. An investigator wants to know if this is also the case in rats.

Strain WKY rats are sometimes used as a model of depression, whereas Wistar rats are not depressive. So he obtains 10 rats of each strain, houses them two per cage for three weeks and tests them in random order using a standard test of pain threshold.

What is the experimental unit in this experiment?       

Each individual rat
Yes. Any two rat strains are 
likely to differ in their pain
threshold. So several depressive
and non-depressive strains need to
be tested to see if sensitivity and
depression are associated
A cage with two
rats in it
The strain of rats
Home Top