22 



THE SPECIFIC HEAT OF WOOD. 



0.340. Between these extremes is a difference of 32 units in the 

 third decimal. It is convenient, then, to group the results into 32 

 classes, those lying between 0.308 and 0.309 forming Class A; those 

 between 0.309 and 0.310 forming Class B, etc. Convenience is the 

 only reason for making 32 classes. Forty or 20 or 25 might as cor- 

 rectly have been formed. Fifty would have been too many, as there 

 would be an average of but two results in each class. Five would 

 have been too few, not because 20 results in each class are too many 

 but because five points would not show all the properties of the curve. 

 The number of results in each one of the 32 classes formed is shown 

 by the ordinates of the successive points on the lowest of the broken 



8\ 



log lot 107 



FIG. 2. Distribution of results through range of initial temperatures, and average results for each half 



lines in figure 3. The sum of the ordinates under this line is 100. 

 The variations in the successive ordinates indicate that the classes 

 have been made too small. The pairing of successive classes AB, 

 BC, CD, etc., and taking the number of observations in each pair, 

 gives the second broken line in which, although the variation of suc- 

 cessive ordinates has been reduced considerably, still further reduc- 

 tion appears advisable. For this purpose successive classes are 

 united in threes, thus ABC, BCD, CDE, etc. When the number of 

 results in each group of three is plotted, the successive frequencies show 

 a much smaller irregular variation, as indicated in the upper broken 



