Slide 1
A control supplies the experiment’s comparison
An experiment tests how a chosen change affects an outcome. A control provides a comparison in which that change is absent or different.
More to Read
Suppose two plant groups receive different fertilizers. If one group also receives more light, the resulting difference cannot be assigned cleanly to fertilizer. The design must address plausible alternative influences.
Slide 2
Define the factors and outcomes before measuring
A factor is something varied or considered in an experiment. Defining factors and outcomes before measurement makes the comparison clearer.
More to Read
Temperature, concentration and duration may interact. Changing only one factor at a time can miss a situation in which a treatment works differently at another temperature; factorial designs investigate combinations.
Slide 3
Random allocation reduces systematic starting differences
Random assignment helps prevent the treatment groups from being systematically different at the start. It does not guarantee identical groups in every experiment.
More to Read
Allocation by convenience can connect treatment with time, location or operator. A randomized run order spreads such influences more fairly, while recording conditions allows remaining patterns to be examined later.
Slide 4
Independent replication needs independent experimental units
Independent replication helps estimate variation. Measuring one specimen repeatedly does not automatically create several independent specimens.
More to Read
Ten readings from one treated container mainly reveal reading variability. Ten independently treated containers address variation between experimental units, which is often the relevant basis for comparing treatments.
Slide 5
The control must fit the question being asked
A strong control is chosen for the specific mechanism being tested. The appropriate comparison depends on what conclusion the experiment should support.
More to Read
A process study may need a standard operating condition, a no-treatment condition, or both. Resources constrain designs, so identifying the decision first helps avoid collecting abundant measurements that cannot answer it.
Sources
- Choosing an Experimental Design
National Institute of Standards and Technology
- Completely Randomized Designs
National Institute of Standards and Technology