Alice Lee 
537 
lOOpr/pi. Taking first England and Wales, and calling these three indices 
respectively /; and 1^, we find : 
Correlation of I4, and li ^ - -9664 ± "0072, 
and Ir = + -9298 + -0148. 
Dr Newshohne gives — '94 as the coefficient of correlation "between segregation 
measured by the fraction of pauper population treated in institutions and the 
phthisis deathrate" (p. 277). Having regard to his confusion of /; and 7^ and 
his frequent interchange of the signs of correlation coefficients, we can only say 
our results confirm his high numerical value, but not his actual figure. 
But does this actual figure mean that there is any real relationship between 
segregation and the phthisis deathrate ? To test this, we replaced the index //, by 
li, where 
r — inn ^^'^^^ number of indoor paupers per lO'* for the population, 1866-1903 
= 100 (f )/(!). 
In this index the relative number of indoor paupers is assumed to remain 
absolutely constant. We found : 
Correlation for England and Wales of 7^ and 7i = — "9459 + "0115, 
that is to say we get substantially the same value, a value higher than 
Dr Newsholaie's, by putting the number of indoor paupers relative to the 
general population constant throughout the period. It is very difficult, in the face 
of such a result, to suppose that segregation of paupers has anything whatever 
to do with the diminution of the phthisis deathrate. It is clearly due to a 
negative correlation of a high magnitude between — ^ and (f>IP, or to a positive 
correlation between ^ and -p, i.e. to a correlation between a high total pauper 
rate and a high phthisis deathrate. Dr Newsholme's result merely reduces to 
the statement that total pauperism in England and Wales has diminished con- 
temporaneously with phthisis. If the result has nothing to do with segregation, 
can we assert that the reduction of phthisis is causally related to the reduction in 
total pauperism ? 
Overlooking for a moment a new objection to be raised later, let us apply the 
variate difference method to the correlation of ^/P with IQOpi/pr and lOOpr/pi 
in the cases of England with Wales, of Scotland, and of Ireland ; also to the 
correlation of (j)/P with the index 100 {pijP)l{p^jP) in the case of England with 
Wales. The following are the results : 
