Unbalanced split-plot designs: Analysis for a non uniform dependence pattern (Q1395971)

Unbalanced designs are quite commonly implemented due to technical problems preventing balanced ones being carried out. The corresponding analysis is performed here for a split-plot design when the usual equicorrelation pattern assumption is removed. We prove that the consequence is a reduction in the number of degrees of freedom in the test function for verifying whether levels of the nested factor are significant. A sensitivity analysis is finally suggested in order to detect whether the reduced number of degrees of freedom is heavily affected by the dependence pattern.