424 
Miscellanea 
It is worth noting that with the exception of one age group (30 — 35) the theoretical frequencies 
corresponding to Professor Edgeworth's curve are consistently worse than both the Type V and 
Type VI frequencies. 
The values of S [e-jU), and S {e-/u) for the three curves are given by the following table: 
TABLE VII. 
S (eyU) 
Type V 
129 
•00053 
•00055 
Type VI 
232 
•00090 
•00099 
Prof. Edgeworth's curve 
868 
•00366 
•00369 
Ncitlicr Type V nor Type VI Hts tlic data well, Lut all the criteria agree in making Professor 
Edgeworth's curve a much i^oorer fit than either of the others. It sliould be observed that in 
order to make com2)arison with Professor Edgeworth's curve possible we have followed hioi in 
grouping all ages above 40 together. This gives a fictitious improvement to the Type V curve, 
for Type VI which depends on the first four moments will fit the "tail" of the experience better 
than Type V whose constants depend on the first three moments only. The value of fj,i is of 
course heavily influenced by the marriages at the higher ages. 
The source of heterogeneity in the spinster curve is misstatement (false or in ignorance) of 
age. I am indebted to Professor Pearson for the remark that Table V shows a lumping up of 
frequency before the decades. If we compare Type VI against the observed, we find a redundancy 
of women just below 30, i.e. at 25, and this is repeated at 45. Again the group 15 — 20 is increased 
by the minors who misstated their age as over 15. Professor Pearson informs me that these 
points have come up for consideration in the treatment of very ample statistics not only English 
but Italian, and in both cases the males give better results than the females. 
We may attempt to circumvent the women who object to entering a new decade by grouping 
in ten year periods round the decades. The following table shows that the removal of this source 
of heterogeneity at once renders Type VI a very good fit, the transition curve Type V a fair fit, 
but leaves Professor Edgeworth's curve a very poor fit. 
TABLE VIII. 
Age 
Observed 
frequency 
Type VI 
Type V 
Piof. Edgeworth's 
curve 
75— 
136088 
136427 
136509 
138329 
35— 
90652 
90298 
90415 
88407 
35— 
7579 
7628 
7462 
45— 
807 
773 
735 
55— 
92 1 Sum 
991 Sum 
1031 Sum 
1 8492 
65— 
34/ =126 
28/ =127 
26/ =129 
Totals 
235252 
235253 
235250 
235228 
We now have = 4-56 when n' = 5 for the Type VI curve so that P = -34 indicating quite 
a good fit. For Type V the value of is H'Ol with n' = 5 giving P= -027, while for Professor 
Edgeworth's curve is 96-31 with n' = 3 so that P is stiU negligible. 
