146 



Viiher^ity of California PuhUcalions in Agricuiiura} Sciences [Vol.4 



If the probable error of the difference of any two percentages in 

 tlie same table is to be obtained, therefore, formulae corresponding to 

 those given by Yule (1911. pp. 264-267) are applicable. 



Now, it is possible in some of these cases to calculate the actual 



standard deviation of the percentage in subsamples which make up 



an aggregate sample. Table 39 gives such actual standard deviations, 



in comparison with the corresponding theoretical or expected standard 



deviations given bv 



pq 



(^ per cent 



s 



3 



n 

 Table 39 



Standard deviations of percentages of mutant types. Values derived from 

 compared with values expressing the actual variability of subsamples. 



PQ, 



° The second values for some cases in this column are derived from p„ (see text). 



