Fig. 162, and a line was drawn on the chart through the different dots representing the factory cost of the machinery. The line was not at all straight and the chief executive spent much time in finding out why there were so many variations from uniformity. He found, among other things, that some of his machinery had not been redesigned for several years, and that the weight of material used was much greater than necessary when taking into account the greater strength of steel and iron made by modern processes. Though materials of the modern kind were being used to a large extent in his machinery, the weight of material had not been reduced and there was more weight of employed material than was actually necessary if new designs were made. Another cause for fluctuation in the curve line was found in the quantities in which the product was manufactured. Some sizes of apparatus were particularly suitable to the public, and on these sizes the quantities were much larger than on other sizes. The sizes more commonly sold were naturally better equipped with jigs and tools than other sizes, and for that reason the cost was lower than would otherwise be expected. After the cost curves had been thoroughly studied for different kinds of apparatus, the selling-price curves were drawn in on the same sheets. It was found that for selling prices, also, there were numerous inconsistencies which could be corrected with advantage to the company. Though some of the peaks and valleys of the selling-price correlation curves were to be expected, there was no justification for others and a little concentrated study brought forth methods by which the selling-price curves could be changed materially with advantage to both the producer and the consumer.
In the study of physics and of experimental engineering, there are many times when a correlation curve is of assistance in the discovery and understanding of the laws of nature. For Fig. 163, many observations were made and recorded on a sheet of co-ordinate paper. After sufficient observations had been made throughout the whole range of the horizontal scale, smooth curves were drawn which would most nearly represent the various dots plotted. In drawing curves of this kind, care should be taken to have each portion of the curve as nearly as possible at the center of gravity of the dots in any vertical section of the chart. Accuracy is not necessarily obtained by having the same number of dots on either side of the curve. If there are only three dots at some vertical line, it may be that two of these dots
