Mixed Effects Strip-Plot Designs under Complete Random Blocks and within-Stage Dependence Pattern

Standard practice with factorial designs or their fractions gives rise to well-known problems. The assumed independence among error terms, linked to constrained randomization of runs, may be no longer accepted, due to grouping of experimental units for economical reasons at each stage of a two--stage treatment process. This paper assumes a general unknown multivariate normal dependence pattern of random components of error terms within each stage due to simultaneous application of factor's $A$ levels within blocks and to simultaneous application of factor's B levels within a common level of factor A. The behaviour of standard testing procedures on effects of a two-factor strip-plot design is examined under these extended assumptions. The main result is that standard strip-plot ANOVA is a robust tool.