18 



Oll thc Sj/fifematic Fitting of Curves 



The ordinates for these curves corresponding to the original observations, i.e. 

 + a-y'/=0, 1/7, 2/7, 3/7, 4/7, 5/7, 6/7, 1, were a.scertained, aiid are given in the 

 accompanying Table. The curves are plotled in Fig. ü. 



Ordinales of Thiele s üb.iei'vations. 



Taking the (Jth parabola as the best let us compare the results found froni it 

 with those obtaiued from the skew curve. Let A, be the difference froui Observation 

 ia the latter case, A., iu the fonner. \Ve find 



New whether we ineasure the g<iodness of fit by the niean A without regard 

 to sign, by the mean square eiror, or by the value of S(A^/i/), we reach the sanie 

 result, — there is an overwhclmiiig balance in favour of the e.xponcutial curve over 

 the algebraic curve. We can make as Thiele* actually does a curve of factorials 

 or even a parabola of the löth order to go through all the 15 observations, but 

 although we shall thus of course get a better fit than by using a three-constant 

 • Forelaesninger over Almindelig lagttageUetUuTe (Ej^benbavn, 1889), p. 12. 



