In the foregoing parts of this thesis the results of the work done have
been considered from two different standpoints, first as they appeared to the
observer without statistical analysis and secondly after critical statistical
methods has been applied to them. Some of the differences which appeared
on first inspection to be significant were not confirmed by the later
statistical analysis. It has been shown that among sheep there are individual
variations of considerable magnitude as well as day -to -day variations depending
on unknown factors beyond experimental control. In fact, many of the difference;
which at first sight appeared to be correlated with diet or season were found
to be explicable as due to the random operation of these individual and day -today variations. The aim of the statistical analysis has, therefore, been to
allow for these two factors and to discover the differences due to other causes.
The discussion is, therefore, limited to those results which have been found
to be "statistically significant".
"Statistically significant" has been defined to dean that a result equally
or more discrepant would be expected to occur less than once in twenty times
as the result of chance. It is, therefore, to be expected that a result of
this type would appear occasionally in the course of a large series of observa:tions even although there was no real heterogeneity. In consequence it is
desirable not to lay too much stress on borderline cases especially if they
seem to be merely isolated occurrences. On the other hand, where a number of
observations each of which is statistically significant, appear to converge
to a general conclusion, such a conclusion way be put forward with more
confidence. It may be added that,. of course, even an isolated result with a
P as low as, say, 0.001, would be given considerable weight but it will be
found that values of P as low as this have been encountered only occasionally
in the present investigation.
In conclusion,it must be clearly recognised that, while evidence has been
elicited that dietary deficiencies affect the natural immunity mechanism, at
the present stage no statement can be made regarding the relative importance
of individual dietary factors. The work done, however, constitutes a step
towards further study of this important question.
These experiments in sheep, however, require to be followed up now by
observations in small animals in which susceptibility to infection can be tested
by direct methods.