The example of shoes designed to last longer, with random assignment and blocking using left and right shoes is not a better design. If the two assignments interact with each other, then both the blocking and the overall experiment fail to achieve their purpose. Suppose having one worn out shoe and one shoe that has little if any wear were to cause discomfort, strain, injury, or aesthetic displeasure. Factors such as these might cause a person to walk less, thereby confounding the experiment. Experiments don't always need "more" blocking, or blocking at lower units of analysis - they need the right level of blocking.
What does he mean by completely randomized design? The assignment to one group or the other wasn't done randomly, it was done evenly between the two groups... (Also, is there such a thing as completely randomized designs?) —Preceding unsigned comment added by 184.108.40.206 (talk) 18:53, 30 October 2007 (UTC)
This page describes stratification, not blocking!