STATISTICAL TREATMENT 



239 



for the control animals, we have 12.5 units, indicating a length of 6.25m 

 to be compared with the value 6.290 obtained by the computation. 

 For the irradiated animals the length as given by the graph is 7.15m 

 to be compared with the computed value of 7.189. To find the standard 

 deviation, we note from Table 3 that deviations in both directions from 

 the mean, of a size greater than one standard deviation, should form 

 31.73 per cent of the observations. Since the distribution is symmetrical, 

 deviations on one side of the mean greater than one standard deviation 

 would constitute one-half this value, or 15.87 per cent. From the graph 



0,01 



0.5 



99.99 



2 10 20 30 40 50 60 70 80 90 98 99.5 



Percentage Occurring Below Indicated Length 



Fig. 3. — Accumulated frequency distribution of lengths of Bodo caudatus plotted on 



arithmetic probability paper. 



for the control group, the percentage 15.87 corresponds to a measurement 

 reading of 10.25, which differs from the reading at the mean (12.5) by 

 2.25 units. Thus the standard deviation obtained graphically is 1.125 m, 

 while that given by the computation was 1.057. For the irradiated 

 animals the difference between the reading at 15.87 and that at 50 per 

 cent is 2.50 units, indicating a standard deviation of 1.25m as compared 

 with the computed value of 1.251. 



At times when we are comparing the difference between means for an 

 experimental and control group we have a series of means such that taking 

 them in corresponding pairs, the experimental group does not differ 

 significantly from the control for any single pair and yet the difference 

 is always in the same direction, which supports the view that a real 

 difference exists. We may accumulate this evidence and test the signifi- 



