154 MEASUREMENT OP HIGH TEMPERATURES. [bull. 54. 
The values of /(0) and/'(0) :/(o) being thus carefully revised it willl 
be expedient to follow the suggestion to which I adverted a moment 
ago, and put 
f(0y(f(0)-f(0)+m)=n (5;, 
By constructing the revised data in Table 34 and taking their graphic 
mean locus I derive the following pairs of correlative values: 
x = 12 y =0.00205 
x' =25 y' =0.00130 
x"=m #"=0.00054 
from which, deducing the constants m and n, 
m=0.000226 w=0.0331 
and the locus /(0) (f(0) :f(0)+m)=n does not differ appreciably fromi 
the mean locus graphically selected. 
Having thus satisfactorily made this preliminary survey it is finally 
desirable to calculate the constants m and n by the method of leasft 
squares. Before doing this equation (5) may be put under a betten 
form for practice by writing 
. /(0}:/(0)=^jy-"» (I 
w here -^t^ is simply the zero value of electrical conductivity of the^ 
alloy whose temperature-coefficient is/'(0) :/(0). Equation (6), whern 
operated on by the method of least squares, does not give inor 
dinate preference to the values /(0) of high resistance, and sinc^ 
the high values can not be warranted with a greater degree of accurac5i 
than the low values, equation (6) may most expediently be made th(* 
basis of computation. This I have done, and the results in Table 41 
contain the observed values of /(0) and f (0) : /(0), the calculated value! J 
/'(0) :/(0), and the errors, and finally the constants m and n, with th(< 
probable errors A„ (m) and A n (n) of each, supposing 
J jr (m)=0.G74 
^(w) = A 
2 (Ayf 
X* 
k-2 
lc2. 
x-(2 
$ 
\2(Ayy 
1c 
X 2 \ xj 
where lc is the number of observations and the factor ( ld= _!l_ ) has « 
0.5 \ 
yfk) 
been suppressed, and where a? and y stand, respectively, for /(0) andti 
f(0) :/(()). The last three alloys, 10, 11, 12, were added subsequently 
to the calculation. 
(308) 
