TRANSACTIONS OF SECTION A. 457 
radium is volatilised at about 1200° C., it would be rendered gaseous by the high 
temperature of the explosion, and redeposited when it cooled. Since the bomb was 
exactly spherical, a change of distribution of the active deposit does not appear 
very probable. In one experiment two electroscopes were used, one by the side of 
the bomb and the other underneath it. Both showed about an equal decrease of 
activity. 
The experiments recorded here are preliminary, and it is intended to examine 
still further whether there is a real change of activity of radium products by the 
action of the high temperature. 
11. On a Freehand Potential Method. By L. F. Ricuarpson. 
This paper described a graphic method for obtaining a vector function of 
position, which function shall satisfy given conditions over a boundary and be 
irrotational and non-divergent everywhere within that boundary. 
In contrast to the usual methods, using Fourier series, Bessel functions, 
spherical or other harmonics, this is very simple. All one has to do is to imitate 
freehand the characteristic common properties of the diagram of intersecting 
stream surfaces and equi-potentials as given by Maxwell, Lamb, Hele-Shaw, and 
others. 
Many people have doubtless used such a method in a rough way, but no one 
seems to have realised the accuracy and power of which it is capable. As far as 
the author has tested it, it will give an accuracy of 1 or 2 per cent. of the 
range considered, which is sufficient for many purposes. 
This method can only be applied when the potential is constant along each 
line of a limited number of families of lines in space. Of these the chief are 
(1) parallel straight lines ; (2) circles in planes normal to, and with their centres 
on, a common axis; (3) radii from a point; (4) the normals to the surfaces of a 
thin shell of any shape ; (5) screw-threads of common and constant pitch about 
& common axis. But within these types the freehand method far surpasses 
analysis in its power of dealing with various forms of surface. 
It will also deal with varying conductivity. It is hoped that it may be of use 
to engineers, 
TUESDAY, AUGUST 6. 
The following Papers and Reports were read :— 
1. Examples of the Modern Methods of treating Observations. 
By W. Pawin Evperton. 
This communication dealt with the application of recognised statistical 
processes to meteorological statistics, The terms mean, median, and mode were 
defined, and the methods of calculating their values given, with examples from 
rainfall statistics. The ‘method of moments’ was described, and as an application 
a parabolic curve was fitted to statistics of rainfall in the east of England. The 
standard deviation was then explained, and it was pointed out that this function 
measures the way the observations are scattered about their mean. The importance 
of calculating a ‘standard deviation’ or a ‘probable error’ corresponding with 
means was then insisted on; the mere statement of a mean is statistically in- 
sufficient and of no use for comparative purposes until we know what deviations 
from the calculated value may arise. 
The remainder of the communication dealt with correlation. The coefficient 
of correlation was first explained, and it was shown how it can be calculated. The 
rainfall in the east of England was compared with that in the north of Scotland 
and in the Channel Islands: in the former case the coefficient of correlation is very 
small, and about equal to its probable error, so that we cannot conclude that there 
is any correlation; but in the latter case the coefficient is found to he -374 + -065, 
