( 408 ) 



Physics. — "Siipplement io Commwiication .\\ 95'^ from the 

 Phi/Mical Lahonitorii of Leiden, on the comparison of the 

 thermo-eJement constantin-steel unth thf hi/drogen thermometer \ 

 By Pi-of. TI. Kamf,rt,ingh Onnks and C. A, Crommelin. 



§ 14. Corrected representation of the oh.'^ervations by a f re term 

 formula. 



As appears from note 1 the calculations in § 12 were made with not 

 perfectly accurate values of the temperature at — 182° and in the 

 same way the mean errors were derived from tlie assumption of 

 those less accurate values. ^) 



If the correct values of those temperatures for the calculations of 

 the deviations W—R„ W~R„ W~R„ W—R^ in Tal)le VIII, are 

 used, the mean errors in microvolts become : 



for formula {Bl) ± 3.0 

 (^11) ±3.4 



{Bill) ± 2.8 (2.5 without —217^) 

 (^IV) ±2.1 

 instead of 



(^I) ± 2.8 



(^11) ± 3.2 



[BIW) ± 2.6 (2.1 without - 217°) 



{B\\) ±1.8 



which would also have been obtained if the observations at — 182° 

 were excluded. 



Now it was necessary to examine whether a repetition of the 

 adjustment would diminish these mean errors. It appeared convincingly 

 that this was not possible to an appreciable degree for {Bl), {B\\), 

 (Z>I1I^. It appeared possible for {B\\') to distribute the errors more 

 equally. However, this only reduced the sum of squares from 26,57 

 to 26,14. 



Instead of the coefficients a^,h^,c^,e^R\\(\f^ (see §12) we get then 



a\ = + 4.32.51 3 e\ = + 0.023276 



b\ = -f 0.4091 53 ƒ , = — 0.0025269 

 c\ = -1-0.0015.563 



The deviations are given in Table IX under W — R\. 



ij The correction amounted to 0°,0S1 in temperature or to 1.7 microvolt, in 

 electromotive force. 



