Blocking

Blocking neutralizes background variables that can not be eliminated by randomizing. It does so by spreading them across the experiment.

You can think of a block as an kind of uncontrolable factor that is added to the experiment. A block is ususally used when this uncontrolable factor cannot be avoided during the experiment, so it is incorporated into the experiment in a controlled way. The idea is to pull the variation due to the blocks out of the expermental error in order to reduce the experimental error and give the test more power.

Common examples of when blocking factors are used:

• When you don’t have enough units from one lot and are forced to use another lot, AND you suspect (maybe know) that there are important differences between lots. You then use (in a controlled manner) half of the units from one lot and half of the units from the other lot.

• When you don’t have enough test chambers for all the parts, AND you suspect (maybe know) that there are important differences in the effects between test chambers. You then assign (in a controlled manner) half of your units to one chamber and half to the other chamber.
• The original use of blocking involved agricultural experiments. Researchers needed to use multiple fields or fields that has important intra-field variation. They would the potential fertility of a field as a block factor. Similar blocking might have been needed due to differences between edge-effects and center effects on crops grown.