l6 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 95 



The last preceding line results from dividing the numbers marked 

 " reciprocals " above into .00125, the reciprocal corresponding to 

 42 (32). 



In a similar manner the computations were made for all 20 wave 

 lengths. After the widely divergent values had been discarded, the 

 mean of all the values of each exposure was then determined, in- 

 cluding all the observed wave lengths found on that exposure. This 

 quantity is called " average multiplier " in the illustration below. 

 Secondly, the products were found of each original virulence value 

 multiplied by the corresponding mean value for each exposure, as 

 shown below by continuing with the illustration of the wave length 

 2463 A. 



Plate no 42 42 46 47 46 47 



Exposure time 32 16 16 16 20 20 



A 1 • 1- z- ^ Virulence 



Average multiplier., t.oo i.oo 1.51 1.29 1.59 1.09 Sum mean 

 2463 A (x 100) 125 .177 .183 .156 .308" .164 .805 .161 



" This value, being wild, is rejected in the mean. 



The mean for each wave length resulting from these final compu- 

 tations constitutes the general result of the whole research on viru- 

 lence for each wave length, and each of the virulence values as plotted 

 against the corresponding wave length gives the curve of radiotoxic 

 virulence shown in figure 2. 



The average deviation of the individual values of virulence from 

 the mean was next determined in the following manner. As shown 

 in the illustration given below, the difiference of the final virulence 

 mean from the individual mean value for each exposure of the in- 

 dividual wave lengths was computed. 



Plate no 42 42 46 47 46 47 a Average Probable 



. ^ ^ ^ Average deviation error 



Exposure time 32 16 16 16 20 20 deviation percent percent 



2463 A (deviation) 36 16 22 5 147" 3 16.4 10.3 3.9 



•■> This value, being wild, is rejected in the mean. 



The mean of these differences for each wave length constitutes the 

 average deviation. The percentage average deviation was then com- 

 puted by dividing the average deviation for each wave length by the 

 corresponding virulence mean. The percentage probable error was 

 next computed according to the formula : 



0.84 percentage average deviation 



where 7? = the number of observations. The percentage probable 

 error is illustrated in fijjure 2. 



