370 
Statistical Study of Anti -Typhoid Inoculation 
r E, 
rJE, 
N 
Table V. 
Transvaal 
-•1504+ -0237 
6-345 
6979 
Table I. 
Stations Abroad ... 
-•0020+ -on 4 
0-175 
31714 
Table VI. 
7 Indian Stations ... 
+ •2395 + •0188 
12-739 
10320 
Table II. 
Indian Stations ... 
+ -2556 + ^0095 
26-905 
40013 
Table III. 
17th Lancers 
+ ^4802 + -0593 
8-098 
689 
Table IV. 
Coldstream Guards 
+ -4987 + -0577 
8-643 
712 
Mean = -2203 ±-0649. 
Mean weighted with r/^,. = -2745 + -0048. 
Mean weighted with jV= -1357 ± -0003. 
We see that four of these results give a positive and two a negative correlation ; 
and the mean calculated with or without weighted ordinates is positive. In 
Tables I, II and V, it will be noted that the number of the inoculated is very 
small compared with the total, and therefore the probable error of the class group — 
inoculated and attacked— is large. A small alteration therefore in this group 
(which is to be reasonably expected in another sample from the general population 
of the inoculated) would consequently largely affect the value of r. A further 
defect in these tables is the very uneven distribution between inoculated and 
uninoculated, which brings the division near the tail of the frequency curve, with 
its disturbing effect on the value of r. From this point of view the most 
satisfactory statistics are found in Tables III, IV and VI, but the total number 
of men included in the two former tables is small. It would appear therefore that 
Table VI (seven large Indian Stations) is the most satisfactory and it has already 
been stated that it is for other reasons the most reliable. In this case the 
coefficient of correlation is sensibly positive and 7'/E,. = 12-739 and is significant. 
The divergent nature of these results calls for some explanation. To some 
extent it may be due to want of homogeneity in the material, owing to different 
methods of inoculation being employed, and to other causes. It does not seem 
probable that the whole effect can be attributed to the results of random sampling 
and so other factors that might influence the result must be sought for. 
There is little doubt that there is more than one disease covered by the term 
Enteric or Typhoid Fever. Bacteriologically it is well known that several varieties 
of para-typhoid bacilli exist and are the specific cause of fevers clinically in- 
distinguishable from typhoid. There is no a priori reason for assuming that a 
vaccine prepared from typhoid bacilli would confer immunity against infection 
with any of the para-typhoid strains, and therefore in a locality where the ratio of 
para- to true typhoid infections is high we might expect to find the coefficient 
of coiTelation between inoculation and freedom from attack correspondingly 
lowered. 
It is, I think, significant that the two highest correlations, i.e. those of the 
17th Lancers and the Coldstream Guards, give very similar results and further 
that both these regiments suffered from well marked epidemics of typhoid, whereas 
