A Tutorial on Randomizing versus Not Randomizing Split-Plot Experiments

This paper studies the impact when the experimenter does not randomize the design for split-plot experiments when there is a linear trend. The paper uses both simulation and basic theory to explain the bias if we do not randomize the design and the inflation in variance if we do. The paper shows proper residual plots can detect the trend if the design is randomized, which allows the analyst to treat time as a covariate, removing the trend's impact. The simulation study provides an excellent way to explain to practitioners the risks from not randomizing the design.