Page:Graphic methods for presenting facts (1914).djvu/142

 tapering form of the horizontal extension of the peak would be understood by even the untrained reader, much more certainly than the chart as shown here with only the figures to indicate the full extent of the loss which occurred in the San Francisco fire.

While Fig. 110 gives some general idea of the proportion which American fire losses bear to the value of new building construction, the two fluctuating curves make it difficult for the reader to make an estimate of the percentage losses year by year. Fig. 111 supplements Fig. 110, and gives for each year the total values for new building construction and the total values of buildings destroyed by fire. Here the percentages of the fire loss are quite obvious when judged by the extent to which the black ink covers the shaded bar representing new building construction. Figures are given in each case for the reader who may care to work out the actual percentage ratios. It must not be assumed from the titles of Fig. 110 and Fig. 111 that the buildings destroyed by fire are the same buildings whose value is recorded in the charts as "new building construction." The rapid advance in the use of fireproof materials makes it likely that the fire losses were more largely from older buildings, built by methods which gave a structure less fireproof than the average for buildings put up in recent years

''Sam. L. Naphtaly, in Journal American Society of Mechanical Engineers''

Fig. 112. Record of Test of a Steam Turbine of 10,000 Kw. Normal Rating at Plant of City Electric Company, San Francisco, California

The different curves shown in this chart supplement each other so as to give all the data on one chart in compact form. The scale for each curve given at the left is only sufficient to show the maximum and minimum value for each curve. The zero lines have been omitted entirely. Though charts of this type with numerous curves are sometimes desirable, they must be carefully made or the reader will be misled by the fluctuations of some one curve appearing more prominently than the data would justify

Fig. 114 shows an error in curve plotting into which it is very easy for an inexperienced person to slip. One vertical scale is relatively