M. Greenwood and J. D. C. White 
515 
Objections may fairly be urged agaiu.st this process. As we have previously remarked, the 
mean is theoretically not the best constant to use for descriptive purposes in the case of 
markedly skew frequencies and we suggested the mode as being more satisfactory. Harvey and 
McKendrick, in an important contribution to the subject*, have alluded to this point and remark 
with justice that the determination of the true mode requires calculation beyond the range of a 
laboratory worker. We have considered whether an approximation of sufficient accuracy might 
not be found. Pearson has shown t that in many cases a good enough value for the mode is 
given by : 
Mode - Median = Twice the Distance from Mean to Median. 
Unfortunately, however, the median cannot be determined with sufficient accuracy in the 
case of discontinuous variates such as phagocytic counts. Thus, we have experimentally no 
finer measurement than the integral number of bacilli per leucocyte ; if we mass in a group all 
the m individuals containing the same number of bacilli and know that the median is the nth 
individual in that group, all we can do is to add of the unit of grouping to the next lower 
integral group measure. This is not nearly close enough in the case of such skew distributions ; 
and we found that the value of the mode thus obtained differed materially from the real mode. 
It was not therefore possible to use this process with success in the case of our .sample curves. 
8S 
Opsonic Index. 
Gkaph 5. Curve of Samples of 100. 
* Biometrika, Vol. vn. p. 64. 
t Pearson, ibid. Vol. i. p. 260. 
