M. Greenwood and J. D. C. White 
TABLE VII. 
519 
Number in Samj^le 
Bi (empirical) 
i>'i (calculated) 
Bo - 3 (empirical) 
B2 - 3 (calculated) 
(20,000 Count) 


25 
•10.33 
•0415 
•4321 
■0505 
50 
•1467 
•0208 
1 ^0907 
■0252 
100 (1st curve) ... 
•3005 
•0104 
1 ^3501 
■0126 
100 (2ud curve) ... 
■3142 
•0104 
1 •4601 
■0126 
(15,000 Count) 
25 
•0457 
•0415 
-•1789 
•0494 
50 
•0980 
•0208 
■7994 
•0247 
100 
•1019 
•0104 
r5418 
•0124 
We can now examine the accuracy of the opsonic method in the way attempted 
in our former paper, viz. we can inquire what the chances are of obtaining from a 
" population " of means, samples giving different indices in terms of the real mean 
of all such samples. In similar cases, it is usual to divide the frequency curve 
into a series of equal areas, setting up ordiuates on either side of the mean or 
mode. Since this elaborate graphical method requires the services of a specially 
skilled draughtsman, and considerably increases the cost of production, we decided 
to adopt the plan followed in our last paper. A considerable number of additional 
ordinates were calculated for each sample curve and we determined the area from 
the beginning or end of the curve up to specified ordinates which corresponded to 
indices of '6, '7, '8, etc. in terms of the mean. From these areas the Tables of 
Chances (VIII, IX, X) were deduced. 
TABLE VIII. 
Chances of Obtaining Certain Deviations in Index Values. 
Opsonic Index in 
terms of the 
Mean 
Samples of 25 
Fraction of the Total 
Area bounded by the 
corresponding ordi- 
nate (Total Area = 1) 
Odds against 
the occurrence of 
such a Deviation 
or a Greater 
■6 
•0089 
111 to 1 
■7 
•0396 
24^3 to 1 
■8 
•1276 
6^8 to 1 
•9 
•2973 
2^4 to 1 
1-1 
•2740 
2^7 to 1 
1-2 
■1324 
6^6 to 1 
1-3 
■0558 
16-9 to 1 
1-fy 
■0211 
46^4 to 1 
Beyond the Hmits I'lt— 
■6 
■0300 
32 ■S to 1 
1-3— 
■7 
■0954 
9^5 to 1 
„ „ 1 
■8 
■2600 
2-9 to 1 
.) I'i— 
■9 
■5713 
■75 to 1 
66—2 
