Rational Methods Of Planning Experiments

In order to find an empirical relation between a quantity F and (m) independent variables or factors, each of which take (n) different values or levels, it is necessary to conduct nm different experiments (all possible combinations of the factors), each of which has to be repeated a few times (usually five) due to the scattering of the data. The total number of such experiments is very large, thus for example, when m = 4 and n = 5, it is equal to 5.54 = 3125; and very rarely is it possible to conduct such a number of experiments. Various possible methods of planning partial experiments are illustrated in Fig. 1. Each method shows a large combinative square consisting of five columns and five rows of middle squares, each of which, in turn, is further divided into five columns and five rows of cells. Such a partition corresponds to a plan of an experiment with four factors, each of which takes five different levels. The blackened cells indicate the combination of factors in a partial experiment. The total number of experiments is reduced on the basis of the following considerations: a) The effect of only a few levels of some factors is studied (Fig. la). Such an investigation gives no information on the effect of some factors, therefore it cannot be accepted as satisfactory; b) By reducing the number of levels of each factor; in such a case it is not possible to obtain an empirical relationship, but only whether some factors have an effect. This is evidently inadequate;