485 
one day but significant variation in all other cases. This variation 
cannot be attributed to smoking. 
The mean values px and py determined during a day cannot be as 
sumed to originate from two distributions which are constant during 
the course of the day. It does not require statistical test to see that the 
variation of px is similar to that of py. Immediately after smoking ope 
rator No 2 had a lower mean value one day while operator No 3 had a 
higher value. Smoking may have had some influence. 
Conclusions 
Only three operators participated and the results are not general. The 
results of this investigation do not indicate that the precision of the 
setting is influenced by tobacco smoking, but the mean values may be. 
The major part of the variation of the mean values is probably due to 
other factors. 
In these experiments the measurements were replicated 25 times to 
form a series. Several series were measured at the same point under 
the same conditions and by the same operator. However, significant 
differences were observed between the mean values of the series. These 
differences may be attributed to several factors, such as irregularities 
of the instrument, play in screws and linkages, temperature variations 
and the human factor in setting and reading. These explain the varia 
tion between the series. The variation within a series can be examined 
by several methods to find significance in settings which are not 
purely random in their variation. These methods include 1) studies of 
the length and number of runs above and below the median, 2) runs 
up and down of successive variations having the same sign, 3) the 
mean square successive difference and others. See reference 11 | 13.3 
13.4 13.5. 
The Mean Square Successive Difference 
The values recorded in a series are denoted x ; , i = 1, 2, 
Accordingly 
25 
x = 2 
1 
Xi 
n 
n — 25 
q 
9 
1 25 
2 (X; - x)2 
11-1 1 
2(n 
24 
— 2 (X; + , — Xj )2 
1) 1 
25.