354 Method of estimating Bacte^^ial Density 



The effect is thus seen to be a fairly uniform one, though distinctly 

 more prominent among the more crowded plates, of which eight pairs of 

 triplets were available. The higher value in the second group is perhaps 

 due to the fact that these contain the counts of the mixed bacterial 

 population in normal milk, while the others are counts of a practically 

 pure culture of B. coli. 



The effect ascribable to "errors of counting" is thus of just the right 

 magnitude to explain the additional variance observed in Series C. Since 

 all the groups are affected similarly and nearly to an equal extent, we may 

 anticipate that if this explanation is correct, the actual values of Series C 

 will fit the theoretical expectation if a uniform allowance of 20 per cent, 

 is made for the additional cause of variation. The distributions are so 

 compared in equal intervals of ^^ in Table XXIII, and by sextiles in 

 Table XXIV. 



Table XXIII 



Table XXIV 



X^ = 7-545, P = -185 (P = -584) 



The distribution shown in Table XXIII shows a remarkably close 

 agreement with expectation. A more exact test of agreement is afforded 

 by the division at the sextiles (Table XXIV); the actual figures show 

 but a moderately good fit with ^ = 7-545, and P = -185; since however 



