stellar dusters 



45 



Denoting the spectral-index by s, the color-index by c we get the following 

 equation of the lines of regression 



(2) 



c = + 0.450 S — 0.027, 

 S = 4- 2.000 C + 0.173. 



The first of these equations gives the mean value of the color-index corres- 

 ponding to a given value of the spectral-index. 



The second equation gives the mean value of the spectral-index corresponding 

 to a given value of the color-index. 



The diagraros fig. 2 and fig. 3 give the geometrical representation of these 

 equations and the observed mean values of 

 c and s for each class. Though systematic 

 deviations are not wanting, we find that^ 

 upon the whole, the observations are well 

 represented by the found regression lines- 



Fig. 2. Mean values of the color-index, correspon- Fig. 3. Mean vaines of the spectral-index, corres- 



ding to given values of the spectral-index. ponding to given values of the color-index. 



The straight line is the regression line, The straight line is the regression line, 



tiie small circles are the observed vakies. the crosses are the observed values. 



According to our assumptions we ought to have c = 0 for s = 0. We may 

 therefore augment the color-indices found by Parkhuest by -(- 0.027. The regres- 

 sion lines now take the form 



(2 



J c — -(- 0.45Ö S, 

 I 5 = + 2.000 c -f 0.119. 



Against the investigation of Parkhurst it may be objected that the determina- 

 tion of the spectral type is somewhat uncertain, owing to the small dimension of 

 the stèliar spectra. This uncertainty is also manifested in the table of Parkhurst, 



