532 
Tuberculosis and Segregation 
the rate of the other is decreasing during the period in question, — always on the 
supposition that we plot the results as Dr Newsholme has done with reversed 
directions of increasing scales for the two indices. He states that " the experience 
is summarised in the high correlation coefficients of '91 for England and Wales 
(1878—1903) and "90 for London (1866—1904)" (p. 271). The correlations 
found from his actual tables do not appear to agree with these, being, for example, 
— "93 for England and Wales with the negative sign as we should anticipate ; but 
as Dr Newsholme does not give the same years for his correlation coefficients as 
in his tables, he may have worked out his coefficients for individual years. It is 
impossible to test the matter, as neither the figures nor their source are provided. 
If, however, we take his Tables LXII and LXIII, and apply the variate 
difference method* to Dr Newsholme's data as they stand in his book, which 
are all the data available, we find 
Correlation of Phthisis Deathrate and Ratio of Deaths in 
Institutions to Total Deaths. 
England and Wales: Third Differences --174 + -293, 
London : Second Differences — '094 + -252. 
In other words the data show no significant relationship between this measure 
of segregation and the phthisis deathrate, when the time-factor is annulled, even 
with the early differences. It is impossible to press the matter further because 
the data are far too sparse for difference treatment, but the results, such as they 
are, are sufficient to indicate that Dr Newsholme's high correlations are solely due 
to the fact that both variates are continuously changing with the timef. 
(3) As a second measure of segregation Dr Newsholme takes lOO^j-/^ and 
1OOO0/'P is then correlated with this, 0^ being the deaths from phthisis in 
institutions. On p. 275 Dr Newsholme gives very meagre data for Brighton, 
Sheffield and Salford in groups of years, six pairs of values for Sheffield, five for 
Brighton and four for Salford. It is thus impossible to test these for annulment 
of the time-factor, and no references are given to the sources of the original data. 
On p. 276 we read : 
Coefficients of correlation summarising this correspondence for long series of single years 
work out at -67 for Salford from 1884 to 1904 and -80 for Sheffield from 1876 to 1905 1. 
If the arithmetical values be correct, they should certainly have negative signs, 
but even then they would not demonstrate anything but the increasing use of 
institutions and the decreasing prevalence of phthisis during the years in question. 
* Biometriha, Vol. x. pp. 179, 341. 
t These values might be modified if we could go to higher differences, but this is impossible on the 
very limited data which Dr Newsholme provides. On these data all we can state is that no evidence 
of organic relationship between the variates, such as is asserted by Dr Newsholme to exist, can be 
demonstrated. 
J There is no statement as to why Brighton has been omitted. 
