rule. This means that with the 10-inch slide rule the accuracy is ordinarily no greater than 1 in 1,000, or one-tenth of one per cent. Though two quantities each running into five figures may be multiplied on the slide rule, the product would not be accurate beyond three significant figures, and ciphers must be put down to express the remainder of the number for the product.
If very large quantities obtained by slide-rule computation are added together with a number of small quantities, the total cannot, of course, be accurate beyond the third or fourth digit toward the right of the largest quantity included in the total. The fourth digit may be fairly accurate in the total, because in the process of addition the various figures added would tend to give a close approximation of the fourth digit and that digit might accordingly be put down in the total because it has at least a fair possibility of accuracy.
It must not be assumed from the preceding paragraph that the slide rule gives figures too crude for ordinary use. There are comparatively few sets of data relating to costs, output, or other records of industrial work which have an accuracy greater than one-tenth of one per cent. For the great majority of ordinary problems, the data are so crude that the 10-inch rule has more than sufficient accuracy. The use of the slide rule on many classes of work has a desirable psychological effect, in that it calls attention to the accuracy of the data and assists in preventing unnecessary detail work which it is very easy to drift into if any assumptions of great accuracy are permitted to creep in.
The question of significant figures in statistical work and even in ordinary commercial reports is an important one which should have greater attention than it ordinarily receives. Unfortunately the subject can be only briefly touched upon here and the reader would do well to look the matter up in some of the books on statistical theory.[1]
It sometimes happens that a few blue prints are required from some complex chart made on heavy paper or cardboard. Instead of making a tracing from the drawing by means of tracing cloth and with great expenditure of labor and time, it is sometimes feasible to treat the original drawing with a transparentizing solution so that blue prints can be obtained directly. The transparentizing solution is put on the paper with a brush or sponge and then blue prints are made in the ordinary manner. There are several different makes of the trans-*
- ↑ A chapter on "Approximation and Accuracy" will be found in "The Elements of Statistical Method," by Willford I. King, published by the Macmillan Company, New York City.
