2022, Vol. 7, Issue 4, Part B

One of the corner stone principles of Design of Experiments is randomization. However, in real life experiments, it is extremely difficult to achieve this. There are settings of the factor levels which are difficult, inconvenient or impossible to change. Moreover, the availability of limited resources makes it difficult to run the entire experiment under homogeneous conditions. These limitations on randomization led to the origin of split-plot designs. These designs were initially used in agricultural experiments but are now extensively being used in industry as they are less expensive, statistically more efficient and give valid results than completely randomized experiments.

In this paper, we have constructed some first-order split-plot designs using non-regular orthogonal designs of order 16. The D-efficiency of the designs for different variance ratios is computed. The designs obtained enable the engineers and scientists to choose the most appropriate columns, for the allocation of Easy-to-change and Hard-to-change factors so that the selected design estimates the required number of parameters with high efficiency.