Designs for Interference

Treatments are often assessed in small experimental units to economize on resources. Treatment differences for the large units used in practice may then be poorly estimated because of interference. Using prior information on treatments, designs are proposed which control interference by ensuring that treatments in neighboring units are similar. Interference may then be ignored in the analysis. Two types of designs are considered: block designs which are based on the concurrence of treatments in the same block, and neighbor designs which take account of the concurrence of treatments in adjacent units.