Split‐Plot Designs

Abstract Split‐plot designs are nested designs that are frequently used in industrial experimentation and result in correlated observations. A correct analysis of the data from split‐plot designs takes account of the correlation. However, as split‐plot designs are often used inadvertently, the typical split‐plot correlation structure is not recognized in many applications. The resulting statistical analysis is then inappropriate. In this contribution, the agricultural origin of split‐plot experimentation is discussed first. Design aspects are discussed as well as the classical analysis of variance (ANOVA) ( see Analysis of Variance ) and industrial applications of split‐plot designs. It is explained that a generalized least‐squares analysis can be used for estimation and inference and different approaches to constructing split‐plot designs are discussed.