Karl Pearson 
The following table gives the observeil and calculated values : 
131 
Microns 
Observed 
Calculated 
Microns 
Observed 
Ualculated 
13 
5 
3-44 
26 
82 
72-74 
8 
5-80 
27 
72 
67 -98 
ir, 
14 
12-25 
28 
50 
59-71 
1 it 
10 
1 / 
iiil 
OO 
17 
40 
37-05 
30 
27 
36-04 
18 
63 
52-87 
31 
' 26 
24-79 
19 
55 
66-68 
32 
18 
15-67 
20 
66 
75-44 
S3 
11 
9-09 
21 
63 
78-43 
34 
4 
4-84 
22 
75 
77-49 
35 
4 
2-37 
28 
87 
75-61 
86 
21^ 
93 
74-71 
|l-75 
25 
80 
74-36 
38 
2 
From these results we find = 29-92 and P — '22. Thus more often than 
once in five trials we should get a worse divergence than the observed, if the 
sample were taken from the calculated population. Some endeavour was made to 
better the fit by small variations from the above solution, discussed by least 
squares, but no improvement was effected. The two components are represented 
in Diagram VII (p. 132). 
(iii) T. gambiense. 
Mean = 22-1130, 
f^, = 14-3389, = 531-3585, 
f^, = 29-1104, /x, = 2429-0948. 
Reducing' nonic : 
24r/ - 7l-7810r/ - 30-o070r/' - 300-0260r/ + 869-6372^^^ 
- 278-847. 5r/ - 270-9547f/- + 58-9108r/ + 14-6050 = 0. 
This leads to p., = - lOr/ = — 9-1777, and the components given in the Table 
p. 125. The two Gaussians and their compound are given in Diagram VIII . 
(p. 133). We find x' = 11-96, giving for n' = 18, P = "80 a splendid fit. 
(iv) Mzimba Strain (from Donkey). 
Mean = 21-4130, 
/x. = 8-7531, ^1^4 = 293-5629, 
fx, = 26-6602, fi, = 1926-7045. 
The reducing nonic : 
24>q-' + 53-5186^/- 25-5876ry''- 41-57062' - 171-2637g^ 
+ 227-1211r/- 37-3371r/-30-8995ry + 8-6177 = 0. 
The required root is p., = ~ lOq = - 4-0000, which leads to the two components 
given in the Table on p. 125. The two components and their c()mi3ound curve are 
17-2 
