Consider Carefully Whether You Need to Block when Designing Experiments

Consider Carefully Whether You Need to Block when Designing Experiments

I think that some experimenters believe that you should always be a case for using blocks when designing experiments. This is not so. There will be circumstances when it will be necessary to consider using blocks, but blocking doesn’t come cheaply. For example, four blocks will require three of your precious degrees of freedom, using up three experimental runs. Furthermore, blocking greatly reduces the range of randomization that can be used. In a blocked experiment randomization of runs can only be carried out within the blocks, instead of across the full set of experimental runs, as is the case in an un-blocked experiment.

Blocking can be useful when you are concerned that factors not included in the design may add variability to the design outcome. Such concern may arise, for example, where you need to divide the experimental runs among several operators, or you need to use two or more different pieces of equipment.

Try to avoid the need for blocking by ensuring that the running of the experiments is carried out with minimum changes among the factors not included in the design. For example, try to have the experiments carried out by just one operator; use the same piece of equipment for all runs; use the same piece of measuring equipment for all measurements from all runs, etc.

Learn more about the correct use of blocking designs by attending our Design of Experiments training courses.