Page:Graphic methods for presenting facts (1914).djvu/353

 There are a great many problems in graphic work which puzzle the person getting up a chart if there are three different variables to deal with. The problem, as ordinarily found, involves two different independent variables, and a dependent variable depending upon each of the two independent variables. Isometric drawings like Fig. 235, or solid models such as are seen in Fig. 236 and Fig. 237, can be used, but they require a great deal of labor and care to make and are accordingly not often seen. There is another method not perhaps so obvious as a solid model but nevertheless of great value. When the data follow any definite natural laws a chart on the style of Fig. 234 is often simple to make and easy to interpret. By such charts many computations may be made with accuracy and ease.

Courtesy of Data, Chicago

Fig. 234. Chart for Obtaining the Weight of Steel Plates 3/8-inch Thick and of Various Widths and Lengths

A chart of this general type permits using three variables. The two independent variables here are length and width. Weight is the dependent variable. This kind of chart is much simpler to construct than an isometric drawing like Fig. 167, or a model like Fig. 236 or Fig. 237

In obtaining curves like those shown in Fig. 234 one of the variables is made some constant quantity, and the other two variables are then used to work out the data from which the curve is drawn. It can be seen, for instance, that if a definite weight and length of steel plate 3/8 inch thick is assumed, the width is absolutely fixed. To obtain a curve like that seen for 5 pounds in Fig. 234, it is necessary only to assume a weight of 5 pounds, then choose separate lengths one by one, and compute the widths which would correspond with the lengths selected to give a weight of 5 pounds. The various figures of width obtained are then plotted as points for the 5-pound curve, and a smooth curve is drawn through all of the points, giving the result seen in the illustration. After one curve has been plotted another weight is assumed in a similar manner, and new computations are made for various lengths on the horizontal scale. Though this method of chart