Experimental Unit — Definition, Formula & Examples
An experimental unit is the individual thing (a person, plant, classroom, or batch) to which a researcher applies a treatment in an experiment. When the units are people, they are usually called subjects.
The experimental unit is the smallest entity to which a treatment is independently and randomly assigned, so that any two distinct units could receive different treatments. It can differ from the observational unit, which is the entity on which the response variable is measured.
Worked Example
Problem: A researcher compares two fertilizers using 8 greenhouses, each holding 20 tomato plants. Fertilizer is delivered through each greenhouse's watering system, so every plant in a greenhouse gets the same fertilizer. She randomly assigns 4 greenhouses to Fertilizer A and 4 to Fertilizer B, then measures the yield of every plant. Identify the experimental unit and the number of units per treatment.
Step 1: Ask what receives the treatment independently. Plants in the same greenhouse cannot receive different fertilizers, so the treatment is assigned to whole greenhouses. The experimental unit is the greenhouse.
Step 2: Count the experimental units and split them across the two treatments.
Step 3: Count the measurements. Each plant is an observational unit, not an experimental unit, because plants were not assigned to treatments one at a time.
Answer: The experimental unit is the greenhouse. There are 8 experimental units, 4 per treatment. The 160 plants are observational units, so the effective sample size for comparing fertilizers is 8, not 160.
Why It Matters
AP Statistics free-response questions often ask you to identify the experimental units before you describe a design, and points are lost when you name the wrong entity. The number of experimental units, not the number of measurements, determines replication and the sample size used in a significance test comparing treatments.
Common Mistakes
Mistake: Treating every measured individual as an experimental unit when treatments were assigned to groups, such as counting 150 students when a teaching method was assigned to 6 whole classrooms.
Correction: Identify the level at which random assignment happened. If whole classrooms were assigned, the classroom is the unit. Measurements within one classroom are not independent replicates, and treating them as such overstates the evidence.