Balanced Experiment

An experiment is balanced when all factor levels (or treatment groups) have the same number of experimental units (or items receiving a treatment). Unbalanced experiments add complexity to the analysis of the data but hopefully for good reason. For example, some levels are of less interest to the researcher than others. Some levels are expected to produce greater variation than others and so more units are assigned to those levels.

Balance is nonessential but desirable if equal accuracy, power, or confidence interval width for treatment comparisons is important. Severe imbalance can induce factor confounding (correlated factors or non-independent treatment levels).