G. MORANT 
323 
These by (xvi) give (3 = •353,862 and equation (xvii) is 
I-196,6514 = log ^ + X -434,2945), 
where z = — , and hence m = 2'580,641. 
VI 
These values of jS and m differ appreciably from those found from the whole 
material ; this possibly being due to the fact that the selected portion was not a 
random sample. The frequency of periods T containing n occurrences is given by 
(xii) and Fig. Ill shows that the theoretical fits the observed distribution quite 
satisfactorily. The P for goodness of fit is "614. 
Fig. III. Number of occurrences in perioils of 10 seconds. 
250 T 
200 
160- 
100 
50- 
(i) - /» =2-581 
' P=-61 
Theory 
(ii) -^m = 2-581 
I P= -47 
- ■ - - shows effect of small change 
in value given to |8 
(iii) Observations 
3 4 5 6 
Number of occurrences 
10 11 
The alternative Xn = log ( ^ ~ . ) method gave 
/3 = -393,772, ??i = 2-495,831, 
and the following distribution. 
For goodness of fit this table gives P= -80 which is a considerable improvement 
on the previous value found {P—'Ql), but the first method is very much less 
