G. MOHANT 
335 
Fig. IV. Lengths of intervals in periods of 10 seconds. 
1000 1 
900 
800 
700 ^ 
600 
I 500-1 
400 
300 
200 
100 
I -I 
I I 
I I 
I I 
yff = -354 
m = 2-581 
P = -88 
OBSERVED 
CALCULATED 
1 
8 
10 
2 3 4 5 6 7 
Leiujthn of intervals 
The /3's deduced from these latter values are so erratic that it is clear that we 
cannot hope to get any reliable value from them. 
These means are shown plotted in Fig. V, thus providing one regression line 
of the surface. The other, giving the mean number of intervals for lengths of 
intervals between t and {t + dt), is calculated theoretically from (xxxi) and it fits 
the observed means quite satisfactorily ; the curve appears to be linear, but we 
have been unable to deduce a proof even of its approximate linearity beyond the 
fact shown above that it is linear for /3 = 0. 
Fig. VI shows the distribution of lengths of intervals in 10 sees, when there 
are three intervals in that period ; one of the theoretical frequencies being that 
calculated by (xxv)'"** while making the total frequency for n - 3 equal to the 
observed. This method would have to be used if only one column of the table were 
known. 
