Karl Pearson 
135 
Reducing nonic : 
24r/ - 224-61152' - 66-6402g'^ - 595-9589f/ + 5079-3305g^ 
-4500-7030r/- 1460-5459f/+ 879-61 IG(/ + 152-2340 = 0. 
The required root is jo., = — lOry = — 4-75085, which leads to the components 
given in Table on p. 125. These components with their compound curve are 
drawn in Diagram X (p. 136). Here ;)j;- = 92-75 which for 20 groups gives 
P< '000,000,1. Thus although the 0. morsitans strain breaks up into two com- 
ponents the combined curve is not a probable description of the frequency. One 
would like to test another sample of this strain, at present it tells against the 
validity of our reduction. 
(vi) Wild-Game Strain. 
Mean = 22-0220, 
/i,= 11-9310, = 404-4932, 
fx, = 29-0514, fji, = 2247-6657. 
Reducing nonic : 
U(f - 18-9446g' - 30-3834f/ - 250-2869f/ + 351-7475r/ 
+ 118-6154g3 - 212-3972f/ + 15-4222r/ + 14-4283 = 0. 
The root required is p., = — lOq = — 6-9859. There result the two components 
provided in the Table p. 125. The two components and their compound are 
figured on Diagram XL (p. 137). We find ;^;== 12-61 giving for n' = 19, P - -81, 
an excellent fit. 
(vii) Humati Strain. 
Mean = 23-7963, 
/x, = 17-0252, 713 1660, 
/X, = 27-1389, IX, = 3034-1222. 
Reducing nonic : 
24f/ - 131-3796r/ - 26-5147f/ - 89-80593-^ + 964-4l76(/ 
- 674-27555^ _ 114-7894g^ + 81-4492(^ + 9-5887 = 0. 
The root is given by p., — — 10q = — 8-557(i, which leads to the components 
given in the Table on p. 125. The two curves and their compound are figured 
in Diagram XII (p. 138). Although the two components merely from the 
graphical point of view do not give a bad fit, the number of trypanosomes in- 
volved is so large that the deviations are not reconcileable with random sampling 
from two such components. We find x' — 79'67, giving P < -000,001. 
